A fault detection method and system for power grid multimedia dispatching system

Abstract

The present invention belongs to the technical field of power grid dispatching, and provides a fault detection and system for a power grid multimedia dispatching system, comprising the following steps: acquiring audio data information of a device and constructing an audio database; extracting modulation spectrum features of the audio database; performing clustering analysis and calculation of the extracted modulation spectrum features of the audio database based on a clustering algorithm to obtain a standard deviation variable; and judging the operating status of the device according to the audio to be tested of the device and the obtained standard deviation variable.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid dispatching, and in particular relates to a fault detection system and a system for a power grid multimedia dispatching system. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the continuous development of social economy and science and technology, more and more electrical equipment appears in people's work and life. While bringing convenience to people, the hidden dangers caused by these electrical equipment to people's lives and property cannot be ignored. In order to avoid or reduce the adverse effects of machine failure, effective monitoring and timely maintenance of the machine are particularly important for the normal operation of the machine.

[0004] At present, machine monitoring generally adopts manual methods, which not only has high maintenance costs and low efficiency, but also may lead to serious failures of the machine not being repaired in time due to inadequate monitoring. In addition, maintenance personnel are an important resource in the maintenance process, but also a scarce resource. However, the existing maintenance personnel management scheme mainly relies on experience, and the personnel scheduling is very random, ignoring the differences in maintenance capabilities of different maintenance personnel, and failing to carry out targeted work.

[0005] In response to the above problems, the maintenance personnel scheduling method based on machine monitoring has received more and more attention. According to the monitored fault types and levels, the maintenance personnel are accurately dispatched to the areas they are good at. This reduces the scheduling cost and improves the maintenance efficiency. Therefore, fault monitoring is particularly important for the scheduling of maintenance personnel. Existing fault monitoring methods usually focus on vibration, pressure and temperature. However, these methods may be too late when the problem is discovered, and the equipment is already in a sub-healthy state or even seriously damaged. Although regular maintenance and repairs can reduce the occurrence of faults to a certain extent, the labor cost is high, the efficiency is low, and the cost performance is low. In recent years, audio monitoring methods have gradually become known to people because of their high efficiency and low cost. Sound-based fault detection and early warning can issue danger warnings in advance so that the machine can be maintained in time.

[0006] Abnormal sound detection is one of the key technologies in audio monitoring methods and can be used for machine maintenance, etc. Abnormal sound detection methods can be divided into supervised and unsupervised methods. Supervised abnormal sound detection (ASD) is generally designed based on the internal structure of the machine and the specific fault sound. There are some related works, but supervised ASD methods are difficult to implement in actual situations. This is because: (1) the frequency of abnormal situations is low and the time of abnormal situation is difficult to predict in advance; (2) it is impractical to deliberately destroy expensive equipment to collect sample data; (3) there are many types of machine failures, and it is difficult to obtain a data set for supervised learning. Another solution (unsupervised ASD method) only requires the sound emitted by the machine when it is operating normally during the training phase. First, a model is created under normal conditions, and then the model is used to calculate the anomaly score of the sampled samples and evaluate the current state. Summary of the invention

[0007] In order to solve the above problems, the present invention proposes a fault detection and system for a power grid multimedia dispatching system. In the training stage, normal sounds are preprocessed to keep them below a certain frequency change threshold, and then the sound modulation spectrum features used to represent the rate of change of signal spectral components over time are extracted. This feature can be used for speech emotion recognition, speech enhancement, etc. Finally, the features are divided into several non-overlapping subsets, each of which is called a cluster. In the detection stage, the features of each sound to be tested are obtained, and the anomaly score is calculated based on the standard-deviation normalized distance between it and the cluster of normal sounds, thereby further improving the accuracy of fault detection.

[0008] According to some embodiments, the first solution of the present disclosure provides a fault detection for a power grid multimedia dispatching system, using the following technical solutions:

[0009] A fault detection method for a power grid multimedia dispatching system includes the following steps:

[0010] Obtain the audio data information of the device and build an audio database;

[0011] Extract modulation spectrum features from audio database;

[0012] Based on the clustering algorithm, cluster analysis and calculation of the modulation spectrum features of the extracted audio database are performed to obtain a standard deviation variable;

[0013] The operating state of the device is determined based on the audio to be tested of the device and the obtained standard deviation variable.

[0014] As a further technical limitation, after acquiring the audio data information of the device, a spectrum gate control method is used to perform noise reduction processing on the audio data information, and an audio database is constructed based on the audio data information after the noise reduction processing.

[0015] As a further technical limitation, the modulation spectrum features of the audio database are extracted through filtering, windowing, discrete Fourier transform and Hilbert transform.

[0016] Furthermore, the filter adopts a triangle filter group and a gammatone filter group.

[0017] As a further technical limitation, clustering is to divide the unlabeled sample data set into several non-overlapping subsets according to certain criteria. Each subset is called a cluster, so that the samples in the same cluster have strong cohesion and the samples in different clusters have obvious separation, thereby revealing the inherent overall structure of the data and discovering the laws and properties contained therein.

[0018] As a further technical limitation, the clustering algorithm adopts a clustering method based on graph theory, that is, the given data information is expressed in the form of a graph, and then the relatively closely linked vertices and related edges in the graph are classified into the same subgraph class to obtain a rough clustering partition graph to achieve clustering.

[0019] As a further technical limitation, in the process of judging the operating status of the device, the audio to be tested is input, the standard deviation normalized distance between the audio to be tested and the standard deviation variable of the cluster analysis result is calculated, and it is judged whether the audio to be tested is abnormal, so as to realize the judgment of the operating status of the device.

[0020] According to some embodiments, the second solution of the present disclosure provides a voiceprint recognition fault monitoring system for power grid dispatching, which adopts the following technical solutions:

[0021] A voiceprint recognition fault monitoring system for power grid dispatching, comprising:

[0022] An acquisition module is configured to acquire audio data information of a device and construct an audio database;

[0023] An extraction module, configured to extract modulation spectrum features of an audio database;

[0024] A clustering module is configured to perform clustering analysis calculation of the extracted audio database modulation spectrum features based on a clustering algorithm to obtain a standard deviation variable;

[0025] The detection module is configured to determine the operating state of the device according to the audio to be tested of the device and the obtained standard deviation variable.

[0026] According to some embodiments, a third solution of the present disclosure provides a computer-readable storage medium, which adopts the following technical solution:

[0027] A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in fault detection for a power grid multimedia dispatching system as described in the first aspect of the present disclosure.

[0028] According to some embodiments, a fourth solution of the present disclosure provides an electronic device, which adopts the following technical solution:

[0029] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in fault detection for a multimedia dispatching system for a power grid as described in the first aspect of the present disclosure are implemented.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. The training set for fault monitoring in the present disclosure only requires normal sound, and the extracted features can capture the long-term dynamic characteristics of the audio signal, highlighting the distinguishability between features.

[0032] 2. In the detection stage of the present invention, an improved graph clustering method is used to capture feature changes, and anomalies are identified by calculating the standard deviation normalized distance, thereby improving the accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0034] Figure 1 is a flowchart of fault detection for a power grid multimedia dispatching system in Embodiment 1 of the present disclosure;

[0035] Figure 2 is a flow chart of extracting modulation spectrum features of an audio database in Embodiment 1 of the present disclosure;

[0036] Figure 3 It is a structural block diagram of the voiceprint recognition fault monitoring system for power grid dispatching in the second embodiment of the present disclosure. DETAILED DESCRIPTION

[0037] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0038] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0039] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0040] In the absence of conflict, the embodiments in the present disclosure and the features in the embodiments may be combined with each other.

[0041] Embodiment 1

[0042] Embodiment 1 of the present disclosure introduces a fault detection method for a power grid multimedia dispatching system.

[0043] like Figure 1 A fault detection method for a power grid multimedia dispatching system is shown, comprising the following steps:

[0044] Step S01: Acquire audio data information of the device and build an audio database;

[0045] Step S02: extracting modulation spectrum features of the audio database;

[0046] Step S03: performing cluster analysis calculation on the extracted audio database modulation spectrum features based on a clustering algorithm to obtain a standard deviation variable;

[0047] Step S04: judging the operating state of the device according to the audio to be tested of the device and the obtained standard deviation variable.

[0048] As one or more implementation modes, in step S01, machine sound audio data is collected, and a spectral gating method is used to perform noise reduction preprocessing on the audio samples;

[0049] Because there are no abnormal sounds in the training data, we first randomly select 100 normal sound clips for each machine and take the average value as noise. This can reduce the impact of other sounds on the monitoring effect except the machine sound. Use Fast Fourier Transform (FFT) to calculate the average power of the noise clip, and calculate the threshold of each frequency band based on the average power. Then calculate the FFT of the normal signal, compare the FFT of the normal signal with the threshold to determine the mask, and smooth it through the filter. Finally, the unnecessary energy changes caused by different loudness levels in the speech signal are eliminated.

[0050] Step S01 specifically includes the following steps:

[0051] Step S101: Divide the data into a training set and a test set according to a certain ratio, wherein the training set contains only mechanical sound and audio samples under normal operation, and the test set contains mechanical sound and audio samples under both normal and abnormal conditions;

[0052] Step S102: Calculate the fast Fourier transform by the noise audio clip;

[0053] Step S103: Obtaining statistical values ​​by performing FFT calculation on the noise;

[0054] Step S104: Calculating a threshold value according to statistical information of noise;

[0055] Step S105: Calculate the FFT of the signal;

[0056] Step S106: Compare the FFT of the signal with the threshold to obtain a mask;

[0057] Step S107: the mask is smoothed by frequency and time filters;

[0058] Step S108: Apply the mask to the spectrogram of the signal.

[0059] As one or more implementation modes, in step S02, Figure 2 As shown in the figure, the modulation spectrum features of the audio sample are extracted through filtering, windowing, discrete Fourier transform and Hilbert transform; the modulation spectrum corresponds to the time-frequency spectrum, which can show the long-term dynamic characteristics of the voiceprint signal; the triangle filter group and the gammatone filter group are often used to extract features of the sound signal. Since the triangle filter group has a poor simulation effect on the basilar membrane of the cochlea, its ability to extract frequency information in a complex environment full of various noises is also relatively poor. The basilar membrane can analyze the frequency of the sound, place different frequency information in the corresponding position, and correspond the size of the frequency energy to the vibration size of the basilar membrane. The gammatone filter group can well simulate the characteristics of the basilar membrane of the human ear.

[0060] Step S02 specifically includes the following steps:

[0061] Step S201: The preprocessed sound signal x(n) is passed through a 60-channel Gammatone filter bank; the first filter of the filter bank is centered at 125Hz, and the last one is below half the sampling rate. The bandwidth follows the equivalent rectangular bandwidth, that is:

[0062]

[0063] Among them, f s is the center frequency of the sth filter, Q earIndicates the quality of the filter at high frequencies, which is set to 9.265 here, B min is the lowest frequency, which is set to 24.7 here;

[0064] Step S202: Perform Hilbert transform on the output of the filter to obtain the time envelope. The calculation process is as follows:

[0065]

[0066] Where n is the time variable, s = 1, 2, 3...60 represents the sth filter, x s (n) is the output of the filter, H{·} is the Hilbert transform, e s (n) is the time envelope. s (n), s = 1, 2, 3 ... 60, add a Hamming window with a window length of 256 ms, and the frame shift is set to 40 ms;

[0067] Step S203: Express the time envelope as e s (m; n) (m is the frame index), and then DFT is performed on it to obtain the modulation spectrum. The process is as follows:

[0068] E s (m; f) = |F(e s (m; n))|

[0069] Where m represents the mth frame obtained after multiplying the Hamming window once, f represents the modulation frequency, F(·) represents the discrete Fourier transform, and E s (m; f) is the modulation spectrum obtained after calculation;

[0070] Step S204: Use a modulation filter bank based on auditory excitation to further group the modulation frequencies into eight bands, denoted as δ (s,k) (m), k = 1...8. Where s is the gammatone filter index, k is the modulation filter index, and m is the frame index; each sound clip gets a 60×8 modulation spectrogram.

[0071] As one or more implementation modes, in step S03, the nearest neighbor graph is used to extract the variation characteristics in the modulation spectrum features of the normal sound audio samples, so as to form clusters of normal sounds, and the mean and standard deviation of each cluster are calculated.

[0072] Clustering is a classic unsupervised learning task that aims to divide an unlabeled sample data set into several non-overlapping subsets according to a certain criterion. Each subset is called a cluster, so that samples in the same cluster have strong cohesion, while samples in different clusters have obvious separation, thereby revealing the inherent overall structure of the data and discovering its underlying laws and properties.

[0073] The basic idea of ​​the clustering method based on graph theory is to express the given data set in the form of a graph, classify the relatively closely linked vertices and related edges in the graph into the same subgraph class, and obtain a rough clustering partition graph to achieve clustering.

[0074] The specific process of step S02 is as follows:

[0075] Consider a graph G, where V is the set of nodes consisting of normal training sound samples and E is the set of edges connecting the nodes. a ,v b ∈V share the same edge set w ab ∈W, and such that:

[0076] v b =argmin({D(v a ,v c )|c=1....a-1,a})

[0077] Among them, D(v a ,v c ) is v a and v c The distance between b is the distance D(v a ,v c )The smallest v c The graph G constructed in this way is composed of several unrelated subgraphs, that is, G = g 1 ∪g 2 ... n ,and Each subgraph is regarded as a separate cluster, and the mean μ corresponding to each cluster is calculated by taking the mean and standard deviation of the modulation spectrogram of the training sound contained in each cluster l and standard deviation σ l .

[0078] As one or more implementations, in step S04, the anomaly score of the sound clip of each machine is calculated using the standard deviation normalized distance of each cluster center corresponding to the machine. Its anomaly score is A t Is It turns out that the process is as follows:

[0079]

[0080] Among them, z t,s,k is the energy value obtained after the t-th test data passes through the s-th gammatone filter and the k-th modulation filter, μ l is the mean of the lth cluster, σl is the standard deviation of the lth cluster.

[0081] Embodiment 2

[0082] The second embodiment of the present disclosure introduces a voiceprint recognition fault monitoring system for power grid dispatching, which adopts the fault detection for power grid multimedia dispatching system introduced in the first embodiment.

[0083] like Figure 3 A voiceprint recognition fault monitoring system for power grid dispatching is shown, comprising:

[0084] An acquisition module is configured to acquire audio data information of a device and construct an audio database;

[0085] An extraction module, configured to extract modulation spectrum features of an audio database;

[0086] A clustering module is configured to perform clustering analysis calculation of the extracted audio database modulation spectrum features based on a clustering algorithm to obtain a standard deviation variable;

[0087] The detection module is configured to determine the operating state of the device according to the audio to be tested of the device and the obtained standard deviation variable.

[0088] The detailed steps are the same as those of the fault detection for the power grid multimedia dispatching system provided in Example 1, and will not be repeated here.

[0089] Embodiment 3

[0090] A third embodiment of the present disclosure provides a computer-readable storage medium.

[0091] A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in fault detection for a power grid multimedia dispatching system as described in the first embodiment of the present disclosure.

[0092] The detailed steps are the same as those of the fault detection for the power grid multimedia dispatching system provided in Example 1, and will not be repeated here.

[0093] Embodiment 4

[0094] A fourth embodiment of the present disclosure provides an electronic device.

[0095] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in fault detection for a multimedia dispatching system for a power grid as described in the first embodiment of the present disclosure are implemented.

[0096] The detailed steps are the same as those of the fault detection for the power grid multimedia dispatching system provided in Example 1, and will not be repeated here.

[0097] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A fault detection method for a power grid multimedia dispatching system, characterized in that: The following steps are involved: Obtain the audio data information of the device and build an audio database; Extract modulation spectrum features from audio database; Based on the clustering algorithm, cluster analysis and calculation of the modulation spectrum features of the extracted audio database are performed to obtain a standard deviation variable; Determine the operating status of the device according to the audio to be tested of the device and the obtained standard deviation variable; The clustering algorithm adopts a clustering method based on graph theory, that is, the given data information is expressed in the form of a graph, and then the relatively closely linked vertices and related edges in the graph are classified into the same subgraph class, and a rough clustering division graph is obtained to achieve clustering; specifically: Consider a graph ,in is a set of nodes consisting of normal training sound samples, is the set of edges connecting nodes, two nodes Share the same edge set , and such that: ； in, yes and The distance between Is able to make distance Smallest The value of; The constructed graph is composed of several unrelated subgraphs, namely ,and , Each subgraph is considered as a separate cluster, and the mean corresponding to each cluster is calculated by taking the mean and standard deviation of the modulation spectrogram of the training sound contained in each cluster and standard deviation .

2. A fault detection method for a power grid multimedia dispatching system as claimed in claim 1, characterized in that: After acquiring the audio data information of the device, a spectrum gate control method is used to perform noise reduction processing on the audio data information, and an audio database is constructed based on the audio data information after the noise reduction processing.

3. A fault detection method for a power grid multimedia dispatching system as claimed in claim 1, characterized in that: The modulation spectrum features of the audio database are extracted through filtering, windowing, discrete Fourier transform and Hilbert transform.

4. A fault detection method for a power grid multimedia dispatching system as claimed in claim 3, characterized in that: The filter adopts a triangle filter group and a gammatone filter group.

5. A fault detection method for a power grid multimedia dispatching system as claimed in claim 1, characterized in that: The clustering is to divide the unlabeled sample data set into several non-overlapping subsets according to certain criteria. Each subset is called a cluster, so that the samples in the same cluster have strong cohesion and the samples in different clusters have obvious separation, thereby revealing the inherent overall structure of the data and discovering the laws and properties contained therein.

6. A fault detection method for a power grid multimedia dispatching system as claimed in claim 1, characterized in that: In the process of judging the operating status of the device, the audio to be tested is input, the standard deviation normalized distance between the audio to be tested and the standard deviation variable of the cluster analysis result is calculated, and it is judged whether the audio to be tested is abnormal, so as to judge the operating status of the device.

7. A voiceprint recognition fault monitoring system for power grid dispatching, which adopts a fault detection method for a power grid multimedia dispatching system as claimed in any one of claims 1 to 6, characterized in that: include: An acquisition module is configured to acquire audio data information of a device and construct an audio database; An extraction module, configured to extract modulation spectrum features of an audio database; A clustering module is configured to perform clustering analysis calculation of the extracted audio database modulation spectrum features based on a clustering algorithm to obtain a standard deviation variable; The detection module is configured to determine the operating state of the device according to the audio to be tested of the device and the obtained standard deviation variable.

8. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the fault detection method for a power grid multimedia dispatching system as described in any one of claims 1 to 6 are implemented.

9. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the fault detection method for a power grid multimedia dispatching system as described in any one of claims 1 to 6 are implemented.

Images