CN116013359A - Power equipment abnormal condition monitoring method and device based on two-stage voiceprint recognition algorithm - Google Patents
Power equipment abnormal condition monitoring method and device based on two-stage voiceprint recognition algorithm Download PDFInfo
- Publication number
- CN116013359A CN116013359A CN202211684287.5A CN202211684287A CN116013359A CN 116013359 A CN116013359 A CN 116013359A CN 202211684287 A CN202211684287 A CN 202211684287A CN 116013359 A CN116013359 A CN 116013359A
- Authority
- CN
- China
- Prior art keywords
- audio data
- power equipment
- stage
- audio
- abnormal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
A method and a device for monitoring abnormal working conditions of electric equipment based on a two-stage voiceprint recognition algorithm are provided, wherein the method comprises the following steps: s1, an audio sensor collects operation audio data of power equipment; s2, preprocessing audio data; s3, enhancing the audio data through a multi-angle mixed data enhancement method; s4, after preprocessing and data enhancement of the audio data, extracting MFCC feature vectors of the audio signals to obtain audio data used for training and/or recognition; s5, during recognition, inputting the audio data extracted by the feature vector into a first-level classification algorithm alpha-ISVM to recognize the abnormal working condition of the power equipment, and entering an abnormal processing flow if the noise is too large or the abnormal working condition of the power equipment is judged to be not learned; otherwise, the power equipment abnormal working condition identification is carried out in the secondary classification algorithm GRU. The method has the beneficial effects of effectively reducing noise interference and having high recognition accuracy.
Description
[ field of technology ]
The invention relates to the technical field of power equipment online monitoring, in particular to a power equipment abnormal condition monitoring method and device based on a two-stage voiceprint recognition algorithm.
[ background Art ]
With the continuous development of the power industry, more and more power equipment is put into each link of a power system, and stable operation of the power equipment is an important guarantee of stable operation of the power system. Monitoring of electrical power equipment is particularly critical because of the particularities of the power industry, where electrical power equipment is often in long-term uninterrupted operation, and any small fault or unstable condition can cause significant losses.
At present, the main monitoring means of the power equipment are more, including monitoring modes such as iron spectrum, vibration, oil chromatography, spectrum, infrared and thermal imaging, and the like, most of the monitoring means are high in cost and large in size, and most of the sensors are required to be placed in the transformer, so that the installation and maintenance are difficult. With the development of artificial intelligence technology, a voiceprint monitoring mode is one of the mainstream monitoring means, and the voiceprint monitoring is used as a non-contact monitoring method, so that the operation of equipment is not interfered, and the installation and maintenance are convenient. However, the existing voiceprint monitoring technology applied in the power industry has the problems of insufficient training data quantity, higher dependence on scenes, poor communication effect with a server, poor equipment integration effect and the like, and greatly influences the recognition accuracy and the monitoring instantaneity.
Edge computing refers to providing near-end services by adopting an open platform with integrated network, computing, storage and application core capabilities on the side close to the object or data source. The application program is initiated at the edge side, and faster network service response is generated, so that the basic requirements of the industry in the aspects of real-time service, application intelligence, security, privacy protection and the like are met. Edge computation is between a physical entity and an industrial connection, or at the top of a physical entity. The cloud computing can still access the historical data of the edge computing.
Incremental learning refers to a learning system that can continually learn new knowledge from new samples and can save a large portion of the knowledge that has been learned before.
Mel (Mel) frequencies are proposed by researchers based on human ear hearing mechanisms and have a non-linear correspondence with hertz (Hz) frequencies. The MFCC (Mel-frequency cepstral coefficients, MFCC for short) calculates the Hz spectrum characteristics by using the nonlinear relation between the two. The computation of the MFCC includes pre-emphasis, framing, windowing, fast fourier transform, mel-filter bank (mel-frequency transform), discrete cosine transform (Discrete Cosine Transform, abbreviated as DCT), dynamic characteristics, and the like.
Raspberry Pi (Chinese name "RaspPi", abbreviated as RPi, or RaspPi/RPI) is designed for learning computer programming education, and only a credit card-sized microcomputer is used, and the system is based on Linux, and with release of Windows 10IoT, the RaspPi running Windows will be generated.
Aiming at the technical problems of insufficient training data quantity and higher dependence on scenes in the prior art of applying a voiceprint monitoring technology in the power industry, the invention technically improves the monitoring method and the device for the abnormal working conditions of the power equipment.
[ invention ]
The invention aims to provide a power equipment abnormal condition monitoring method which effectively reduces noise interference and has high identification accuracy.
In order to achieve the above purpose, the technical scheme adopted by the invention is a power equipment abnormal condition monitoring method based on a two-stage voiceprint recognition algorithm, which comprises the following steps:
s1, an audio sensor collects operation audio data of power equipment;
s2, preprocessing the audio data, reducing noise interference through end point cutting, audio mixing, normalization processing, pre-emphasis, framing and windowing, and enhancing feature expression;
s3, enhancing the audio data by a multi-angle mixed data enhancement method, enhancing the audio data by adding reverberation, characteristic time shift, pitch correction and waveform stretching from three angles of tone, loudness and quality according to the characteristics of the audio signal, and reducing the specificity of the audio data without changing the structural characteristics of the audio data;
s4, after preprocessing and data enhancement of the audio data, extracting MFCC feature vectors of the audio signals to obtain audio data used for training and/or recognition;
s5, during recognition, inputting the audio data extracted by the feature vector into a first-level classification algorithm alpha-ISVM to recognize the abnormal working condition of the power equipment, and entering an abnormal processing flow if the noise is too large or the abnormal working condition of the power equipment is judged to be not learned; otherwise, the power equipment abnormal working condition identification is carried out in the secondary classification algorithm GRU.
Preferably, the method for monitoring abnormal working conditions of the power equipment based on the two-stage voiceprint recognition algorithm further comprises the following steps:
and S6, during training, an incremental learning algorithm is realized through a forgetting factor alpha in a first-stage classification algorithm alpha-ISVM, in the process of gradually acquiring new samples of the audio data, partial elements in the old samples are eliminated according to the forgetting factor alpha according to the weight of the samples, the classification model is optimized while the accuracy of the subsequent classifier is ensured, and the storage space is reduced.
Preferably, step S6: after new samples of the audio data are collected, the sample distribution is optimized by using an incremental learning algorithm, a forgetting factor alpha=0.6 is selected, partial old samples are selectively forgotten, and the classification recognition model is retrained.
Preferably, step S3: and adding Gaussian white noise with the mean value of 0 and the standard deviation of 1 to increase reverberation, shifting partial features to positive directions by half a period, setting the pitch correction step length to 3, and generating more general feature data.
Still another object of the present invention is to provide an abnormal condition monitoring apparatus for electrical equipment, which effectively reduces noise interference and has high recognition accuracy.
In order to achieve the above object, the present invention provides a power equipment abnormal condition monitoring device based on a two-stage voiceprint recognition algorithm, which is an edge computing device developed based on an embedded system motherboard, and is used for executing the above power equipment abnormal condition monitoring method based on the two-stage voiceprint recognition algorithm.
Preferably, the device for monitoring the abnormal working condition of the power equipment based on the two-stage voiceprint recognition algorithm comprises a 5G communication module, wherein the 5G communication module is used for storing the collected audio information and fault information, is connected with a server in a working scene of the power equipment, and realizes front-end and back-end communication.
Preferably, the device for monitoring abnormal working conditions of electric equipment based on the two-stage voiceprint recognition algorithm is developed based on a raspberry group 4B main board, an audio signal is collected and used by a ReSPe αker2-Mics PiH A T microphone, and a 5G communication module is realized by adopting an SIM820X-M2 5G H A T extension version.
The method and the device for monitoring the abnormal working condition of the power equipment based on the two-stage voiceprint recognition algorithm have the following beneficial effects: the automatic monitoring and recognition of the voiceprints of the power equipment are realized, and the operation threshold and the distribution cost are reduced; the method has the advantages that the audio data enhancement and sample automation iteration are realized, the operation is simplified, the training speed is reduced, and the model identification accuracy is improved; based on data enhancement and two-stage classification algorithm, the voice print-based power equipment abnormal condition monitoring process has a 5G communication function, and the voice print-based power equipment abnormal condition monitoring process is integrated into a set of equipment, so that the equipment has the functions of classifying and identifying, returning identification information and communicating with a server, and meanwhile, the equipment has self-optimizing capability in the using process by utilizing algorithm optimization.
[ description of the drawings ]
FIG. 1 is a flow chart of a method for monitoring abnormal working conditions of power equipment based on a two-stage voiceprint recognition algorithm.
FIG. 2 is a schematic structural diagram of an abnormal working condition monitoring device of an electric power device based on a two-stage voiceprint recognition algorithm.
[ detailed description ] of the invention
The invention is further described below with reference to examples and with reference to the accompanying drawings.
Example 1
The embodiment realizes a power equipment abnormal condition monitoring method based on a two-stage voiceprint recognition algorithm.
The method of the embodiment is based on data enhancement and two-stage classification algorithm and has a 5G communication function.
FIG. 1 is a flow chart of a method for monitoring abnormal working conditions of power equipment based on a two-stage voiceprint recognition algorithm. As shown in fig. 1, the method of the embodiment is based on an audio sensor, an edge computing device, a 5G communication module, a multi-angle mixed data enhancement algorithm, an alpha-ISVM, a GRU two-stage classification algorithm and an incremental learning algorithm.
The audio sensor collects the running audio data of the power equipment, transmits the running audio data to the edge computing equipment, processes the audio data by utilizing a multi-angle mixed data enhancement algorithm, and transmits the audio data to an alpha-ISVM and GRU two-stage classification algorithm for recognition. And uploading the acquired audio data and the identification result to a server through a 5G module. And taking the newly acquired audio data as a new sample, and fine-tuning the model by using an incremental learning algorithm.
The multi-angle mixed data enhancement method is an algorithm for enhancing audio data, and according to the characteristics of audio signals, the audio data is enhanced by adding reverberation, characteristic time shift, pitch correction and waveform stretching from three angles of tone, loudness and quality, the structural characteristics of the data are not changed, the specificity of the data is reduced, and excessive fitting during model training is avoided, so that the model is more universal.
The alpha-ISVM and GRU two-stage classification algorithm is a combination of two classification algorithms, the alpha-ISVM is a support vector machine classification algorithm with an incremental learning function, and the GRU is a cyclic neural network. After preprocessing and data enhancement, the audio data still contains noise components, after the audio is extracted by the MFCC feature vector, the audio data is input into a first-level algorithm alpha-ISVM for recognition, and if the noise is too large or is judged to be of an unlearned class, the abnormal processing flow is entered. And otherwise, entering a second-level algorithm GRU for identification. Compared with a single classification algorithm, the two-stage algorithm can effectively reduce noise interference and improve recognition accuracy.
The incremental learning algorithm is an algorithm for optimizing sample space distribution and selectively forgetting to eliminate training data, is realized through forgetting factors alpha in alpha-ISVM, eliminates part of elements in old samples according to the weight of the samples according to the forgetting factors alpha in the process of gradually acquiring new samples, optimizes the model while ensuring the precision of subsequent classifiers, and reduces the occupation of storage space.
The 5G communication module is a communication component which can be used for edge computing equipment, is connected with a server in a working scene of the power equipment through the module, realizes front-end and back-end communication and is used for storing collected audio information and fault information.
The embodiment realizes the automatic monitoring and recognition of the voiceprint of the power equipment, and reduces the operation threshold and the distribution cost; the method has the advantages of realizing audio data enhancement and sample automation iteration, simplifying operation, reducing training speed and improving model identification accuracy.
Example 2
The embodiment realizes a power equipment abnormal condition monitoring device based on a two-stage voiceprint recognition algorithm, which is used for executing the power equipment abnormal condition monitoring method based on the two-stage voiceprint recognition algorithm described in embodiment 1.
FIG. 2 is a schematic structural diagram of an abnormal working condition monitoring device of an electric power device based on a two-stage voiceprint recognition algorithm. As shown in fig. 2, in this embodiment, the raspberry group 4B main board is selected as a development board of the voiceprint monitoring device, and the audio signal is collected and collected by using a reppe αker 2-ics Pi H a T microphone and is sent to the development board for audio preprocessing. Noise interference is reduced by endpoint clipping, audio mixing, normalization processing, pre-emphasis, framing and windowing, and feature expression is enhanced.
The preprocessed data is further processed through a multi-angle data enhancement method, gaussian white noise with the mean value of 0 and the standard deviation of 1 is selected to be added for reverberation, partial features are shifted to the positive direction for half a period, the pitch correction step length is set to be 3, more general feature data are generated, MFCC feature vectors are extracted, and audio data which can be used for training and recognition are obtained.
The 5G communication function of the device is realized by adopting an SIM820X-M2 5G H A T extension version, communication is supported by adopting an SIM card, communication with a back-end server is realized, and data exchange is completed.
And taking the processed audio data as an identification object, transmitting the identification object into a first-stage algorithm alpha-ISVM for first classification identification, entering an abnormal processing flow if the noise is too large or the noise is judged to be an unlearned class, alarming at a development board end, uploading the identification object to a server through a 5G module, and recording abnormal information. If the result is not strange, the result is input into a secondary algorithm GRU for secondary classification and identification, a single-layer GRU is selected to prevent overfitting of the overmodel, the identification result is output to a display interface externally connected with a raspberry group development board, and the result information is uploaded to a server through a 5G module for backup.
After new samples are collected, the sample distribution is optimized by using an incremental learning algorithm, a forgetting factor alpha=0.6 is selected, part of old samples are selectively forgotten, and the recognition model is retrained, so that the new recognition model which is more suitable for the current working condition is obtained.
According to the embodiment, the abnormal working condition monitoring flow of the electric equipment based on the voiceprint is integrated in one set of equipment, so that the equipment has the functions of classifying and identifying, returning identification information and communicating with a server, and meanwhile, the equipment has the self-optimizing capability in the using process by utilizing algorithm optimization.
It will be appreciated by those of ordinary skill in the art that all or part of the steps of implementing the above embodiments may be implemented by hardware, or may be implemented by a program to instruct related hardware, where the program may be stored in a computer readable storage medium, where the storage medium may be a magnetic disk, an optical disc, a Read-Only Memory (ROM), a random access Memory (Random Acess Memory, RAM), or the like.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and additions to the present invention may be made by those skilled in the art without departing from the principles of the present invention and such modifications and additions are to be considered as well as within the scope of the present invention.
Claims (7)
1. The power equipment abnormal condition monitoring method based on the two-stage voiceprint recognition algorithm is characterized by comprising the following steps of:
s1, an audio sensor collects operation audio data of power equipment;
s2, preprocessing the audio data, reducing noise interference through end point cutting, audio mixing, normalization processing, pre-emphasis, framing and windowing, and enhancing feature expression;
s3, enhancing the audio data by a multi-angle mixed data enhancement method, enhancing the audio data by adding reverberation, characteristic time shift, pitch correction and waveform stretching from three angles of tone, loudness and quality according to the characteristics of the audio signal, and reducing the specificity of the audio data without changing the structural characteristics of the audio data;
s4, after preprocessing and data enhancement of the audio data, extracting MFCC feature vectors of the audio signals to obtain audio data used for training and/or recognition;
s5, during recognition, inputting the audio data extracted by the feature vector into a first-level classification algorithm alpha-ISVM to recognize the abnormal working condition of the power equipment, and entering an abnormal processing flow if the noise is too large or the abnormal working condition of the power equipment is judged to be not learned; otherwise, the power equipment abnormal working condition identification is carried out in the secondary classification algorithm GRU.
2. The method for monitoring abnormal conditions of power equipment based on the two-stage voiceprint recognition algorithm according to claim 1, further comprising the following steps:
and S6, during training, an incremental learning algorithm is realized through a forgetting factor alpha in a first-stage classification algorithm alpha-ISVM, in the process of gradually acquiring new samples of the audio data, partial elements in the old samples are eliminated according to the forgetting factor alpha according to the weight of the samples, the classification model is optimized while the accuracy of the subsequent classifier is ensured, and the storage space is reduced.
3. The method for monitoring abnormal working conditions of the electrical equipment based on the two-stage voiceprint recognition algorithm as claimed in claim 2, wherein the step S6 is: after new samples of the audio data are collected, the sample distribution is optimized by using an incremental learning algorithm, a forgetting factor alpha=0.6 is selected, partial old samples are selectively forgotten, and the classification recognition model is retrained.
4. The method for monitoring abnormal working conditions of the electrical equipment based on the two-stage voiceprint recognition algorithm according to claim 1, wherein the method is characterized by comprising the following step S3: and adding Gaussian white noise with the mean value of 0 and the standard deviation of 1 to increase reverberation, shifting partial features to positive directions by half a period, setting the pitch correction step length to 3, and generating more general feature data.
5. An abnormal working condition monitoring device of power equipment based on two-stage voiceprint recognition algorithm is characterized in that: the method is edge computing equipment developed based on an embedded system main board and is used for executing the method for monitoring the abnormal working condition of the power equipment based on the two-stage voiceprint recognition algorithm according to any one of claims 1 to 4.
6. The device for monitoring abnormal conditions of electrical equipment based on the two-stage voiceprint recognition algorithm according to claim 5, wherein the device is characterized in that: the system comprises a 5G communication module, wherein the 5G communication module is used for storing collected audio information and fault information, and is connected with a server in a working scene of power equipment to realize front-end and back-end communication.
7. The device for monitoring abnormal conditions of electrical equipment based on the two-stage voiceprint recognition algorithm according to claim 6, wherein the device is characterized in that: based on the development of the raspberry pie 4B main board, the audio signal acquisition is realized by selecting a ReSPe alpha ker2-Mics PiH A T microphone and the 5G communication module is realized by adopting an SIM820X-M2 5G H A T extension version.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211684287.5A CN116013359A (en) | 2022-12-27 | 2022-12-27 | Power equipment abnormal condition monitoring method and device based on two-stage voiceprint recognition algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211684287.5A CN116013359A (en) | 2022-12-27 | 2022-12-27 | Power equipment abnormal condition monitoring method and device based on two-stage voiceprint recognition algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116013359A true CN116013359A (en) | 2023-04-25 |
Family
ID=86034825
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211684287.5A Pending CN116013359A (en) | 2022-12-27 | 2022-12-27 | Power equipment abnormal condition monitoring method and device based on two-stage voiceprint recognition algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116013359A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117233589A (en) * | 2023-09-14 | 2023-12-15 | 中国南方电网有限责任公司超高压输电公司广州局 | GIS equipment fault diagnosis method and device, computer equipment and storage medium |
| CN117894317A (en) * | 2024-03-14 | 2024-04-16 | 沈阳智帮电气设备有限公司 | Box-type transformer on-line monitoring method and system based on voiceprint analysis |
-
2022
- 2022-12-27 CN CN202211684287.5A patent/CN116013359A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117233589A (en) * | 2023-09-14 | 2023-12-15 | 中国南方电网有限责任公司超高压输电公司广州局 | GIS equipment fault diagnosis method and device, computer equipment and storage medium |
| CN117894317A (en) * | 2024-03-14 | 2024-04-16 | 沈阳智帮电气设备有限公司 | Box-type transformer on-line monitoring method and system based on voiceprint analysis |
| CN117894317B (en) * | 2024-03-14 | 2024-05-24 | 沈阳智帮电气设备有限公司 | Box-type transformer on-line monitoring method and system based on voiceprint analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116013359A (en) | Power equipment abnormal condition monitoring method and device based on two-stage voiceprint recognition algorithm | |
| JP6792089B2 (en) | Voiceprint recognition methods, devices, terminal devices and storage media | |
| CN109256150B (en) | Speech emotion recognition system and method based on machine learning | |
| US5412738A (en) | Recognition system, particularly for recognising people | |
| CN108962231A (en) | A kind of method of speech classification, device, server and storage medium | |
| CN111326178A (en) | Multi-mode speech emotion recognition system and method based on convolutional neural network | |
| CN112562698B (en) | Power equipment defect diagnosis method based on fusion of sound source information and thermal imaging characteristics | |
| CN111145763A (en) | GRU-based voice recognition method and system in audio | |
| CN116490920A (en) | Method for detecting an audio challenge, corresponding device, computer program product and computer readable carrier medium for a speech input processed by an automatic speech recognition system | |
| CN113823293A (en) | Speaker recognition method and system based on voice enhancement | |
| CN110970020A (en) | Method for extracting effective voice signal by using voiceprint | |
| CN112017658A (en) | Operation control system based on intelligent human-computer interaction | |
| CN111091840A (en) | Method for establishing gender identification model and gender identification method | |
| CN115910097A (en) | Audible signal identification method and system for latent fault of high-voltage circuit breaker | |
| WO2021217979A1 (en) | Voiceprint recognition method and apparatus, and device and storage medium | |
| CN113077812A (en) | Speech signal generation model training method, echo cancellation method, device and equipment | |
| CN116741159A (en) | Audio classification and model training method and device, electronic equipment and storage medium | |
| CN111951793B (en) | Method, device and storage medium for awakening word recognition | |
| CN114121025A (en) | Voiceprint fault intelligent detection method and device for substation equipment | |
| KR20220012473A (en) | Apparatus and method unsupervised pretraining speaker embedding extraction system using mutual information neural estimator, computer-readable storage medium and computer program | |
| CN111914777B (en) | Method and system for identifying robot instruction in cross-mode manner | |
| CN111933187B (en) | Emotion recognition model training method and device, computer equipment and storage medium | |
| CN113689863A (en) | Voiceprint feature extraction method, device, equipment and storage medium | |
| Vasuki et al. | Emotion recognition using ensemble of cepstral, perceptual and temporal features | |
| CN118116391A (en) | Robust voiceprint-based anti-fraud system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |