CN114325214A - Electric power online monitoring method based on microphone array sound source positioning technology - Google Patents
Electric power online monitoring method based on microphone array sound source positioning technology Download PDFInfo
- Publication number
- CN114325214A CN114325214A CN202111372611.5A CN202111372611A CN114325214A CN 114325214 A CN114325214 A CN 114325214A CN 202111372611 A CN202111372611 A CN 202111372611A CN 114325214 A CN114325214 A CN 114325214A
- Authority
- CN
- China
- Prior art keywords
- signal
- microphone array
- sound source
- microphone
- signals
- 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
- Circuit For Audible Band Transducer (AREA)
Abstract
The invention belongs to the technical field of electric power systems and automation thereof, and particularly relates to an electric power online monitoring method based on a microphone array sound source positioning technology. The invention comprises the following steps: based on a genetic algorithm, acquiring signals by adopting a non-uniform microphone array; preprocessing the acquired signals; extracting the characteristics of the preprocessed signals; classifying and identifying the extracted signal characteristics through a neural network, and constructing a convolutional neural network; and outputting the identified signal characteristics. The invention provides a whole set of solution for online monitoring of power equipment, including sound collection positioning, signal feature extraction and neural network classification and identification, aiming at the use condition of the power equipment and combining with digital energy network development planning. The sound and the environmental noise of other surrounding equipment can be effectively removed, the device is suitable for being used in environments such as a transformer substation, the requirements of a digital transformer network can be met, and the device has a positive effect on accelerating the development of a digital power grid.
Description
Technical Field
The invention belongs to the technical field of electric power systems and automation thereof, and particularly relates to an electric power online monitoring method based on a microphone array sound source positioning technology.
Background
At present, in the rapid popularization of the digital twin technology of the power network in China, more power online monitoring methods are required to be applied to the power network, and no online monitoring technology for effectively positioning the operation and fault sound of power equipment is available.
Therefore, a complete set of solutions for online monitoring of power equipment, including sound collection and positioning, signal feature extraction, and neural network classification and identification, become a new subject continuously developed by those skilled in the art.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides an electric power on-line monitoring method based on a microphone array sound source positioning technology. The method aims to realize the invention aims of a whole set of solution method for realizing the on-line monitoring of the power equipment, including sound acquisition and positioning, signal feature extraction and neural network classification and identification.
The technical scheme adopted by the invention for realizing the purpose is as follows:
an electric power on-line monitoring method based on a microphone array sound source positioning technology is realized by a microphone array signal acquisition system and comprises the following steps:
and 5, outputting the identified signal characteristics.
Further, the genetic algorithm process comprises the following steps:
step (1) initializing a cluster;
step (2), evaluating the isolation fitness inside the cluster;
selecting;
step (4) crossing;
step (5), mutation;
step (6) the iteration condition is terminated;
and (7) finishing the step.
Further, the evaluating the intra-cluster isolation fitness in the step (2) comprises:
randomly selecting 32 points and 32 fixed points in the area to form a sample, generating a population in a circulating manner, wherein the range of horizontal and vertical coordinates of the population sample is [ -256, 256] mm, and the generated coordinate is rounded to 0; if the repeated coordinates appear, regenerating new random coordinates, and circulating until no repeated coordinates exist; evaluating a fitness objective function, wherein the higher the objective function is, the higher the selection probability of the sample is, and the calculation steps are as follows:
a. calculating individual fitness x (x) in the populationi) (i ═ 1,2, …, N), N being the number of individuals in the population;
b. genetic Secondary probability P (x)i) Cumulative population probability Q
The selecting of the step (3) comprises:
in [0, 1 ]]Generating S uniformly distributed pseudo-random numbers T in intervalsIf T iss<Q1Selecting a population 1; if Qm-1-<T<QmSelecting a population m, repeating S times to select S groups of populations to enter a new sample;
the step (4) of interleaving comprises:
randomly selecting M groups of populations to carry out sample crossing (M < N and even number) in pairs, selecting M (M <32) random coordinates in each group to carry out crossing operation, and calculating the method as follows:
A1＝αA(1)+(1-α)A(2)
A2＝(1-α)A(1)+αA(2)
in the above formula, A1And A2Two groups of new population coordinate points, A, generated after the cross operation is finished(1)And A(2)Two groups of population coordinate points selected for cross operation, alpha is a generated random number between (0,1), and the generated coordinates are rounded to 0;
the mutation of the step (5) comprises the following steps:
setting a small fixed probability value PVBeta is a random number in the interval (0,1), if beta < PVThen mutation occurs;
and (3) coordinate variation operation: a. theV＝βA+(1-β)Z
Wherein A isVIs a regenerated coordinate point, Z is a randomly generated coordinate point in the array interval range, and A is a randomly selected coordinate point in the variant population;
the iteration condition termination of step (6) comprises:
the iteration condition is terminated according to the set difference of the target functions of the two generations of samples, and the default is less than 5%.
Further, the step 2 of preprocessing the acquired signals includes:
pre-emphasis of signals is performed to make up for the loss of high-frequency components in the audio signal acquisition process, so that a flatter frequency spectrum is obtained; pre-emphasis high-pass filter H (z) 1-alphaz-1And alpha is normal, and is in the interval of (0.9-1.0).
Further, the step 3 of extracting the features of the preprocessed signals includes extracting position features and extracting sound features; the sound feature extraction is to calculate the load reduction for the neural network by identifying short-time energy, zero crossing rate, main harmonic set and power spectrum density characteristics and based on the time domain characteristics and frequency domain characteristics of the signals; the algorithm of the sound source positioning system comprises a band-pass filtering unit, a windowing detection unit, an endpoint detection unit, a signal-to-noise ratio evaluation unit, a Fourier transform unit, a cross power spectrum unit, a frequency domain weighting unit, an inverse Fourier transform unit, a peak detection unit and a geometric positioning unit.
Further, the algorithm of the sound source localization system comprises the following steps:
step (1) band-pass filtering;
firstly, filtering a signal, designing a band-pass filter, and passing the signal at a frequency of 200hz-4000 hz;
step (2) windowing and detecting;
the signal is divided into frames with the length of 10-30ms (sampling points 1024), and the frames are 1/2 frames;
step (3) end point detection and Fourier transform;
carrying out endpoint detection and Fourier transform (DFT and FFT) on the framed signals, wherein the endpoint detection mainly calculates whether short-time energy and short-time average zero crossing rate reach a threshold value or not, and judges whether the sound source signals are effective active sounds or not;
step (4) signal-to-noise ratio estimation;
after the endpoint detection program judges that the signal is an effective signal, entering a signal-to-noise ratio estimation program;
judging the current frame signal S according to the end point detection partk(n) whether it is a valid signal, when it is a valid signal, by
To estimate Sk(n) prior signal-to-noise ratio, where E (k-1) is the short-term energy of the non-significant signal of the nearest frame to the current frame, to obtain the current frame Sk(n) SNR is α SNRk+ (1- α) SNR (k); when S isk(n) inactive signal, update E (k-1) ═ En(k) To prepare for the next frame speech signal to obtain the signal-to-noise ratio, update the current frame SkThe signal-to-noise ratio of (n) is SNR (k) ═ SNR (k-1); storing the SNR value obtained finally into an appointed address memory so as to be called by a frequency domain weighting module;
step (5), cross-power spectrum;
after Fourier transformation, signals enter a cross-power spectrum module, wherein the signals are X + jY and C + jS respectively, and a cross-power spectrum result R + jI is (X + jY) (C + jS);
step (6) frequency domain weighting;
the signal is weighted in frequency domain through cross power spectrum, and the characteristic is that the weighting function changes along with the change of signal-to-noise ratio; adopting color gold PHAT-GCC algorithm
Step (7), inverse Fourier transform;
carrying out inverse Fourier transform on the cross-power spectrum western number after the frequency domain weighting to obtain a cross-correlation function;
step (8) peak value detection;
step (9) geometric positioning;
estimating the relative position of a sound source in space according to the relative time delay between microphone arrays, and obtaining a positioning formula:
furthermore, the step 4 of classifying and identifying the extracted signal features through a neural network to construct a convolutional neural network, wherein the convolutional neural network is constructed through the structures of an input map, a convolutional layer, a pooling layer, a normalization layer, a full-link layer and an output layer; the method comprises the following steps:
the method comprises the steps of classifying and sampling operation and fault sounds of power equipment such as a transformer, a mutual inductor, a switch and a bus, wherein each class of 1000-segment audio forms original sample data;
the length of each audio frequency is about 0.2-20s, the sampling rate of the audio data is 44.1KHz, the sampling bit number is 16 bits, part of the data is dual-channel sampling, and the audio format is WAV;
dividing the audio signal samples into frames, wherein the frame length is 256 points, the frame shift is 128 points, a single sample is divided into 32 frames, and an audio characteristic block with 32 continuous frames is constructed; extracting mixed characteristic parameters by taking each frame as a unit, and taking 12-order MFCC parameters, 12-order MFCC first-order difference coefficients, 12-order GFCC parameters and 12-order GFCC first-order difference coefficients as input layers;
extracting the characteristics of the data of the input layer through the convolution layer, wherein the convolution kernel is 3 multiplied by 3, the moving step length is 1, and the activation function is a tanh function;
the data output by the convolutional layer passes through a pooling layer, the pooling mode is maximum pooling, and the size of a pooling domain is 2 multiplied by 2; adding a normalization layer before the pooled data is used as an input parameter for the next set of convolutional layers to increase local competition between adjacent features;
and the output layer uses a Softmax regression algorithm to map the obtained characteristics to two classification results of whether the power equipment sounds, and substituting the MFCC and the first-order difference thereof, the GFCC and the first-order difference thereof, and the MFCC and GFCC characteristic parameters into the identification network respectively as input layers.
Furthermore, the microphone array signal acquisition system comprises a microphone spherical array, a terminal, a power supply and an upper computer PC; the power supply supplies power to the microphone array and the terminal, the microphone array is connected with the terminal through an audio line, and the terminal is communicated with the upper computer through a USB interface; the terminal comprises a signal input interface, a main control development board, a flash memory and a data output interface.
Furthermore, the microphone array signal acquisition system is provided with a spoke-shaped or 3D spherical microphone array and comprises a microphone two-channel driving circuit, wherein the connection relationship of the microphone two-channel driving circuit is as follows:
the dual-channel circuit needs a processor to provide 3I/O ports, a CLK port is used as a clock signal input of a microphone, a WS port is used for selecting a channel when the microphone works, and a dual-channel common sound DATA output port DATA is arranged; by changing the frequency of the clock input terminal SCK, the sampling rate of the microphone can be changed, and meanwhile, the power consumption and the working mode of the microphone are controlled.
A computer storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the steps of a method for online monitoring of electrical power based on a microphone array sound source localization technique.
The invention has the following beneficial effects and advantages:
the invention designs a whole set of solution schemes for online monitoring of the power equipment, such as sound acquisition and positioning, signal feature extraction and neural network classification and identification, aiming at the use condition of the power equipment and combining the development and planning of the current digital energy network in China.
The basic principle of the scheme is that a beam forming algorithm is adopted based on a microphone array, and finally an enhanced signal pointing to a target direction is obtained (beam forming). The super-strong pointing capability of the beam forming technology can effectively remove the sound and environmental noise of other surrounding equipment, so that the method is very suitable for being used in environments such as transformer substations and the like. The invention meets the requirements of a digital power transformation network and has a positive effect on accelerating the development of a digital power grid.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a work flow diagram of the method of the present invention;
FIG. 2 is a flow chart of the genetic algorithm of the present invention;
FIG. 3 is an external view of a microphone array of the present invention;
FIG. 4 is a two channel circuit diagram of the present invention;
FIG. 5 is a block diagram of the algorithmic design of the sound source localization system of the present invention;
fig. 6 is a diagram of a CNN network architecture according to the present invention;
fig. 7 is a schematic view of the monitoring device of the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
The solution of some embodiments of the invention is described below with reference to fig. 1-7.
Example 1
The invention provides an embodiment, which is an electric power on-line monitoring method based on a microphone array sound source localization technology, as shown in fig. 1, fig. 1 is a work flow chart of the method of the invention.
The invention relates to a wheat husk wind array design scheme based on a genetic algorithm, which is designed through a microphone array signal acquisition system design scheme, a sound source positioning system algorithm and a power equipment sound feature identification method based on a convolutional neural network.
The method specifically comprises the following steps:
by utilizing a genetic algorithm, the microphone array with uniform layout is optimized to generate grating lobes more easily under the condition of a certain size of the microphone array, and the problem of airspace confusion is solved. Therefore, the present invention adopts a non-uniform microphone arrangement, as shown in fig. 2, and fig. 2 is a flow chart of the genetic algorithm of the present invention.
The genetic algorithm process comprises the following steps:
step (1) initializing a cluster;
step (2), evaluating the isolation fitness inside the cluster;
and randomly selecting 32 points and 32 fixed points in the area to form a sample, circularly generating a population, wherein the range of the horizontal and vertical coordinates of the population sample is [ -256, 256] mm, and the generated coordinates are rounded to 0. And if repeated coordinates appear, regenerating new random coordinates, and circulating until no repeated coordinates exist. Then, a fitness objective function is evaluated, the higher the objective function is, the higher the selection probability of the sample is, and the calculation steps are as follows:
a. calculating individual fitness x (x) in the populationi) (i-1, 2, …, N), N being the number of individuals in the population.
b. Genetic Secondary probability P (x)i) The population cumulative probability Q;
selecting;
in [0, 1 ]]Generating S uniformly distributed pseudo-random numbers T in intervalsIf T iss<Q1Selecting a population 1; if Qm-1-<T<QmThen selecting population m, repeating S times to select S groups of populations to enter new sample
Step (4) crossing;
the M groups of populations were randomly selected for sample crossing (M < N, and even). And m (m <32) random coordinates are selected for each group to carry out cross operation, and the calculation mode is as follows:
A1＝αA(1)+(1-α)A(2)
A2＝(1-α)A(1)+αA(2)
in the above formula, A1And A2Two groups of new population coordinate points, A, generated after the cross operation is finished(1)And A(2)Are two groups of population coordinate points selected for the crossover operation, α is the generated random number between (0,1), and the generated coordinates are rounded to 0.
Step (5), mutation;
firstly, a smaller fixed variation probability value P is setVBeta is a random number in the interval (0,1), if beta < PVThen it is changedAnd (3) distinguishing.
And (3) coordinate variation operation: a. theV＝βA+(1-β)Z；
Wherein A isVIs a regenerated coordinate point, Z is a randomly generated coordinate point within the array interval range, and a is a randomly selected coordinate point in the variant population.
Step (6) the iteration condition is terminated;
the iteration condition is terminated according to the set difference of the target functions of the two generations of samples, and the default is less than 5%.
And (7) finishing the step.
The microphone array signal acquisition system is provided with a spoke-shaped or 3D spherical microphone array, the structure is shown in figure 3, and figure 3 is an appearance diagram of the microphone array of the invention.
The microphone two-channel driving circuit diagram of the microphone array signal acquisition system is shown in fig. 4, the circuit needs a processor to provide at least 3 IO ports, a BCLK port is used as a clock signal input of a microphone, a WS port is used for selecting a channel when a pair of microphones work, and a sound DATA output port DATA. The sampling rate of the microphone can be changed by changing the frequency of the clock input end SCK, the power consumption and the working mode of the microphone are controlled simultaneously, and a microphone array circuit is formed by a plurality of microphone driving circuits.
The connection relation of the driving circuit is as follows:
a binaural circuit requires a processor to provide 3I/O ports, a CLK port as a clock signal input to the microphone, a WS port for channel selection when a pair of microphones is operated, and a binaural common sound DATA output port DATA. By changing the frequency of the clock input terminal SCK, the sampling rate of the microphone can be changed, and meanwhile, the power consumption and the working mode of the microphone are controlled.
the pre-processing is mainly pre-emphasis of the signal. The signal pre-emphasis is to compensate for the loss of high frequency components during the audio signal acquisition process, thereby obtaining a flatter frequency spectrum. Pre-emphasis high-pass filter H (z) 1-alphaz-1And alpha is normal, and is in the interval of (0.9-1.0).
the feature extraction includes position feature extraction and sound feature extraction. Fig. 5 is a block diagram showing the algorithm design of the sound source localization system of the present invention, as shown in fig. 5. And carrying out position feature extraction on the preprocessed signals through an algorithm of a sound source positioning system. The sound feature extraction is to calculate the load reduction for the neural network by identifying the characteristics of short-time energy, zero crossing rate, main harmonic set, power spectral density and the like and based on the time domain characteristics and the frequency domain characteristics of the signals.
The algorithm of the sound source positioning system is composed of a band-pass filtering unit, a windowing detection unit, an end point detection unit, a signal-to-noise ratio evaluation unit, a Fourier transform unit, a cross power spectrum unit, a frequency domain weighting unit, an inverse Fourier transform unit, a peak detection unit and a geometric positioning unit.
The algorithm of the sound source localization system comprises the following steps:
and (1) band-pass filtering.
Firstly, filtering a signal, designing a band-pass filter, and passing the signal at a frequency of 200hz-4000 hz;
and (2) windowing and detecting.
The signal is divided into frames with the length of 10-30ms (sampling points 1024), and the frames are 1/2 frames;
and (3) detecting an end point and performing Fourier transform.
And performing endpoint detection and Fourier transform (DFT and FFT) on the framed signal, wherein the endpoint detection mainly calculates whether the short-time energy and the short-time average zero-crossing rate reach a threshold value, and judges whether the sound source signal is effective active sound.
And (4) estimating the signal-to-noise ratio.
After the endpoint detection program judges that the signal is an effective signal, entering a signal-to-noise ratio estimation program;
judging the current frame signal S according to the end point detection partk(n) whether it is a valid signal, when it is a valid signal, by
To estimate Sk(n) a priori signal-to-noise ratio, where E (k-1) is the invalid signal from the nearest frame to the current frameShort-time energy of the number to obtain the current frame Sk(n) SNR is α SNRk+ (1- α) SNR (k). When S isk(n) inactive signal, update E (k-1) ═ En(k) To prepare for the next frame speech signal to obtain the signal-to-noise ratio, update the current frame SkThe signal-to-noise ratio of (n) is SNR (k) ═ SNR (k-1). And storing the SNR value obtained finally into a designated address memory so as to be called by a frequency domain weighting module.
And (5) cross-power spectrum.
The signals enter a cross-power spectrum module after Fourier transformation, the signals are respectively X + jY and C + jS, and a cross-power spectrum result R + jI is (X + jY) (C + jS).
And (6) weighting the frequency domain.
The signals are weighted in frequency domain through cross power spectrum, and the characteristic is that the weighting function changes along with the change of the signal-to-noise ratio. Adopting color gold PHAT-GCC algorithm
And (7) performing inverse Fourier transform.
And performing inverse Fourier transform on the cross-power spectrum west number after the frequency domain weighting to obtain a cross-correlation function.
And (8) peak value detection.
And (9) geometrically positioning.
Estimating the relative position of a sound source in space according to the relative time delay between microphone arrays, and obtaining a positioning formula:
specifically, a convolutional neural network is constructed through the structures of an input map, a convolutional layer, a pooling layer, a normalization layer, a full-link layer and an output layer. As shown in fig. 6, fig. 6 is a diagram of a CNN network structure of the present invention.
In specific implementation, the method is realized by the following scheme:
the operation and fault sound of power equipment such as a transformer, a mutual inductor, a switch, a bus and the like are classified and sampled, and each class of 1000-segment audio forms original sample data.
The length of each audio is about 0.2-20s, the sampling rate of the audio data is 44.1KHz, the sampling bit number is 16 bits, part of the data is binaural sampling, and the audio format is WAV.
And (3) dividing the audio signal samples into frames, wherein the frame length is 256 points, the frame shift is 128 points, the single sample is divided into 32 frames, and an audio characteristic block with 32 continuous frames is constructed. And extracting mixed characteristic parameters by taking each frame as a unit, and taking a 12-order MFCC parameter, a 12-order MFCC first-order difference coefficient, a 12-order GFCC parameter and a 12-order GFCC first-order difference coefficient. It is used as an input layer.
And then extracting the characteristics of the input layer data through the convolution layer. The convolution kernel is 3 × 3, the step size of the move is 1, and the activation function is a tanh function.
The data output by the convolutional layer passes through the pooling layer, the pooling mode is maximum pooling, and the size of a pooling area is 2 multiplied by 2. A normalization layer is added before pooling data is used as input parameters for the next set of convolutional layers, increasing the local competition between adjacent features.
And the output layer uses a Softmax regression algorithm to map the obtained characteristics to two classification results of whether the power equipment sounds, and substituting the MFCC and the first-order difference thereof, the GFCC and the first-order difference thereof, and the MFCC and GFCC characteristic parameters into the identification network respectively as input layers.
And 5, outputting the identified signal characteristics.
Example 2
The invention provides an embodiment, which is an electric power on-line monitoring method based on a microphone array sound source localization technology, and the method is realized by an electric power on-line monitoring device based on the microphone array sound source localization technology, as shown in fig. 7, and fig. 7 is a schematic structural diagram of the monitoring device of the invention.
The monitoring equipment comprises a microphone array signal acquisition system hardware part, a microphone spherical array, a terminal machine, a power supply and an upper computer PC. The power supply supplies power to the microphone array and the terminal machine, the microphone array is connected with the terminal machine through an audio line, and the terminal machine is communicated with the upper computer through a USB interface. The terminal comprises a signal input interface, a main control development board, a flash memory and a data output interface.
Example 3
Based on the same inventive concept, an embodiment of the present invention further provides a computer storage medium, where a computer program is stored on the computer storage medium, and when the computer program is executed by a processor, the steps of the method for monitoring power online based on a microphone array sound source localization technology described in embodiment 1 are implemented.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (10)
1. An electric power on-line monitoring method based on a microphone array sound source positioning technology is characterized in that: the method is realized by a microphone array signal acquisition system and comprises the following steps:
step 1, acquiring signals by adopting a non-uniform microphone array based on a genetic algorithm;
step 2, preprocessing the acquired signals;
step 3, extracting the characteristics of the preprocessed signals;
step 4, classifying and identifying the extracted signal characteristics through a neural network, and constructing a convolutional neural network;
and 5, outputting the identified signal characteristics.
2. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 1, wherein: the genetic algorithm process comprises the following steps:
step (1) initializing a cluster;
step (2), evaluating the isolation fitness inside the cluster;
selecting;
step (4) crossing;
step (5), mutation;
step (6) the iteration condition is terminated;
and (7) finishing the step.
3. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 2, wherein: the evaluation of the isolation fitness inside the cluster in the step (2) comprises the following steps:
randomly selecting 32 points and 32 fixed points in the area to form a sample, generating a population in a circulating manner, wherein the range of horizontal and vertical coordinates of the population sample is [ -256, 256] mm, and the generated coordinate is rounded to 0; if the repeated coordinates appear, regenerating new random coordinates, and circulating until no repeated coordinates exist; evaluating a fitness objective function, wherein the higher the objective function is, the higher the selection probability of the sample is, and the calculation steps are as follows:
a. calculating individual fitness x (x) in the populationi) (i ═ 1,2, …, N), N being the number of individuals in the population;
b. genetic Secondary probability P (x)i) Cumulative population probability Q
The selecting of the step (3) comprises:
in [0, 1 ]]Generating S uniformly distributed pseudo-random numbers T in intervalsIf T iss<Q1Selecting a population 1; if Qm-1-<T<QmSelecting a population m, repeating S times to select S groups of populations to enter a new sample;
the step (4) of interleaving comprises:
randomly selecting M groups of populations to carry out sample crossing (M < N and even number) in pairs, selecting M (M <32) random coordinates in each group to carry out crossing operation, and calculating the method as follows:
A1＝αA(1)+(1-α)A(2)
A2＝(1-α)A(1)+αA(2)
in the above formula, A1And A2Two groups of new population coordinate points, A, generated after the cross operation is finished(1)And A(2)Two groups of population coordinate points selected for cross operation, alpha is a generated random number between (0,1), and the generated coordinates are rounded to 0;
the mutation of the step (5) comprises the following steps:
setting a small fixed probability value PVBeta is a random number in the interval (0,1), if beta < PVThen mutation occurs;
and (3) coordinate variation operation: a. theV＝βA+(1-β)Z
Wherein A isVIs a regenerated coordinate point, Z is a randomly generated coordinate point in the array interval range, and A is a randomly selected coordinate point in the variant population;
the iteration condition termination of step (6) comprises:
the iteration condition is terminated according to the set difference of the target functions of the two generations of samples, and the default is less than 5%.
4. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 1, wherein: step 2, preprocessing the acquired signals, comprising:
pre-emphasis of signals is performed to make up for the loss of high-frequency components in the audio signal acquisition process, so that a flatter frequency spectrum is obtained; pre-emphasis high-pass filter H (z) 1-alphaz-1And alpha is normal, and is in the interval of (0.9-1.0).
5. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 1, wherein: step 3, performing feature extraction on the preprocessed signals, wherein the feature extraction comprises position feature extraction and sound feature extraction; the sound feature extraction is to calculate the load reduction for the neural network by identifying short-time energy, zero crossing rate, main harmonic set and power spectrum density characteristics and based on the time domain characteristics and frequency domain characteristics of the signals; the algorithm of the sound source positioning system comprises a band-pass filtering unit, a windowing detection unit, an endpoint detection unit, a signal-to-noise ratio evaluation unit, a Fourier transform unit, a cross power spectrum unit, a frequency domain weighting unit, an inverse Fourier transform unit, a peak detection unit and a geometric positioning unit.
6. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 4, wherein: the algorithm of the sound source localization system comprises the following steps:
step (1) band-pass filtering;
firstly, filtering a signal, designing a band-pass filter, and passing the signal at a frequency of 200hz-4000 hz;
step (2) windowing and detecting;
the signal is divided into frames with the length of 10-30ms (sampling points 1024), and the frames are 1/2 frames;
step (3) end point detection and Fourier transform;
carrying out endpoint detection and Fourier transform (DFT and FFT) on the framed signals, wherein the endpoint detection mainly calculates whether short-time energy and short-time average zero crossing rate reach a threshold value or not, and judges whether the sound source signals are effective active sounds or not;
step (4) signal-to-noise ratio estimation;
after the endpoint detection program judges that the signal is an effective signal, entering a signal-to-noise ratio estimation program;
judging the current frame signal S according to the end point detection partk(n) whether it is a valid signal, when it is a valid signal, by
To estimate Sk(n) prior signal-to-noise ratio, where E (k-1) is the short-term energy of the non-significant signal of the nearest frame to the current frame, to obtain the current frame Sk(n) SNR is α SNRk+ (1- α) SNR (k); when S isk(n) inactive signal, update E (k-1) ═ En(k) Evaluating the SNR for the next frame of speech signalPreparing, updating the current frame SkThe signal-to-noise ratio of (n) is SNR (k) ═ SNR (k-1); storing the SNR value obtained finally into an appointed address memory so as to be called by a frequency domain weighting module;
step (5), cross-power spectrum;
after Fourier transformation, signals enter a cross-power spectrum module, wherein the signals are X + jY and C + jS respectively, and a cross-power spectrum result R + jI is (X + jY) (C + jS);
step (6) frequency domain weighting;
the signal is weighted in frequency domain through cross power spectrum, and the characteristic is that the weighting function changes along with the change of signal-to-noise ratio; adopting color gold PHAT-GCC algorithm
Step (7), inverse Fourier transform;
carrying out inverse Fourier transform on the cross-power spectrum western number after the frequency domain weighting to obtain a cross-correlation function;
step (8) peak value detection;
step (9) geometric positioning;
estimating the relative position of a sound source in space according to the relative time delay between microphone arrays, and obtaining a positioning formula:
7. the on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 1, wherein: step 4, classifying and identifying the extracted signal features through a neural network to construct a convolutional neural network, wherein the convolutional neural network is constructed through the structures of an input map, a convolutional layer, a pooling layer, a normalization layer, a full-link layer and an output layer; the method comprises the following steps:
the method comprises the steps of classifying and sampling operation and fault sounds of power equipment such as a transformer, a mutual inductor, a switch and a bus, wherein each class of 1000-segment audio forms original sample data;
the length of each audio frequency is about 0.2-20s, the sampling rate of the audio data is 44.1KHz, the sampling bit number is 16 bits, part of the data is dual-channel sampling, and the audio format is WAV;
dividing the audio signal samples into frames, wherein the frame length is 256 points, the frame shift is 128 points, a single sample is divided into 32 frames, and an audio characteristic block with 32 continuous frames is constructed; extracting mixed characteristic parameters by taking each frame as a unit, and taking 12-order MFCC parameters, 12-order MFCC first-order difference coefficients, 12-order GFCC parameters and 12-order GFCC first-order difference coefficients as input layers;
extracting the characteristics of the data of the input layer through the convolution layer, wherein the convolution kernel is 3 multiplied by 3, the moving step length is 1, and the activation function is a tanh function;
the data output by the convolutional layer passes through a pooling layer, the pooling mode is maximum pooling, and the size of a pooling domain is 2 multiplied by 2; adding a normalization layer before the pooled data is used as an input parameter for the next set of convolutional layers to increase local competition between adjacent features;
and the output layer uses a Softmax regression algorithm to map the obtained characteristics to two classification results of whether the power equipment sounds, and substituting the MFCC and the first-order difference thereof, the GFCC and the first-order difference thereof, and the MFCC and GFCC characteristic parameters into the identification network respectively as input layers.
8. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 1, wherein: the microphone array signal acquisition system comprises a microphone spherical array, a terminal, a power supply and an upper computer PC; the power supply supplies power to the microphone array and the terminal, the microphone array is connected with the terminal through an audio line, and the terminal is communicated with the upper computer through a USB interface; the terminal comprises a signal input interface, a main control development board, a flash memory and a data output interface.
9. The on-line power monitoring method based on the microphone array sound source localization technology as claimed in claim 1, wherein: the microphone array signal acquisition system is provided with a spoke-shaped or 3D spherical microphone array and comprises a microphone two-channel drive circuit, and the connection relation of the microphone two-channel drive circuit is as follows:
the dual-channel circuit needs a processor to provide 3I/O ports, a CLK port is used as a clock signal input of a microphone, a WS port is used for selecting a channel when the microphone works, and a dual-channel common sound DATA output port DATA is arranged; by changing the frequency of the clock input terminal SCK, the sampling rate of the microphone can be changed, and meanwhile, the power consumption and the working mode of the microphone are controlled.
10. A computer storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of a method for online monitoring of electrical power based on microphone array sound source localization techniques as claimed in claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111372611.5A CN114325214A (en) | 2021-11-18 | 2021-11-18 | Electric power online monitoring method based on microphone array sound source positioning technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111372611.5A CN114325214A (en) | 2021-11-18 | 2021-11-18 | Electric power online monitoring method based on microphone array sound source positioning technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114325214A true CN114325214A (en) | 2022-04-12 |
Family
ID=81046635
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111372611.5A Pending CN114325214A (en) | 2021-11-18 | 2021-11-18 | Electric power online monitoring method based on microphone array sound source positioning technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114325214A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115061086A (en) * | 2022-05-12 | 2022-09-16 | 上海事凡物联网科技有限公司 | Moving target detection method based on micro-aperture microphone array |
| CN115169250A (en) * | 2022-09-05 | 2022-10-11 | 湖北工业大学 | Microphone array optimization method and system based on improved mixed integer differential evolution |
| CN115662436A (en) * | 2022-11-14 | 2023-01-31 | 北京探境科技有限公司 | Audio processing method and device, storage medium and intelligent glasses |
| CN117172135A (en) * | 2023-11-02 | 2023-12-05 | 山东省科霖检测有限公司 | Intelligent noise monitoring management method and system |
| CN117289208A (en) * | 2023-11-24 | 2023-12-26 | 北京瑞森新谱科技股份有限公司 | Sound source positioning method and device |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102854494A (en) * | 2012-08-08 | 2013-01-02 | Tcl集团股份有限公司 | Sound source locating method and device |
| CN106709216A (en) * | 2017-02-24 | 2017-05-24 | 中国航天空气动力技术研究院 | Method for optimally designing microphone array by taking acoustic propagation correlation loss into consideration |
| CN106792415A (en) * | 2016-12-26 | 2017-05-31 | 歌尔科技有限公司 | A kind of method of testing and device for digital microphone array |
| CN110221996A (en) * | 2019-05-16 | 2019-09-10 | 北京小米移动软件有限公司 | Controller configuration method and device in main control chip |
| CN110335617A (en) * | 2019-05-24 | 2019-10-15 | 国网新疆电力有限公司乌鲁木齐供电公司 | A kind of noise analysis method in substation |
| CN111883161A (en) * | 2020-07-08 | 2020-11-03 | 东方通信股份有限公司 | Method and device for audio acquisition and position identification |
| CN112183647A (en) * | 2020-09-30 | 2021-01-05 | 国网山西省电力公司大同供电公司 | Transformer substation equipment sound fault detection and positioning method based on deep learning |
| CN212519401U (en) * | 2020-06-22 | 2021-02-09 | 深圳市鸿合创新信息技术有限责任公司 | Digital microphone circuit and pickup electronic equipment |
| CN112348052A (en) * | 2020-09-30 | 2021-02-09 | 国网山西省电力公司大同供电公司 | Power transmission and transformation equipment abnormal sound source positioning method based on improved EfficientNet |
| CN112710987A (en) * | 2020-12-28 | 2021-04-27 | 哈尔滨理工大学 | Microphone array system for identifying and positioning illegal whistle vehicle |
| CN112906795A (en) * | 2021-02-23 | 2021-06-04 | 江苏聆世科技有限公司 | Whistle vehicle judgment method based on convolutional neural network |
| CN113176537A (en) * | 2020-01-24 | 2021-07-27 | 动态Ad有限责任公司 | Detection and classification of siren signals and location of siren signal source |
-
2021
- 2021-11-18 CN CN202111372611.5A patent/CN114325214A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102854494A (en) * | 2012-08-08 | 2013-01-02 | Tcl集团股份有限公司 | Sound source locating method and device |
| CN106792415A (en) * | 2016-12-26 | 2017-05-31 | 歌尔科技有限公司 | A kind of method of testing and device for digital microphone array |
| CN106709216A (en) * | 2017-02-24 | 2017-05-24 | 中国航天空气动力技术研究院 | Method for optimally designing microphone array by taking acoustic propagation correlation loss into consideration |
| CN110221996A (en) * | 2019-05-16 | 2019-09-10 | 北京小米移动软件有限公司 | Controller configuration method and device in main control chip |
| CN110335617A (en) * | 2019-05-24 | 2019-10-15 | 国网新疆电力有限公司乌鲁木齐供电公司 | A kind of noise analysis method in substation |
| CN113176537A (en) * | 2020-01-24 | 2021-07-27 | 动态Ad有限责任公司 | Detection and classification of siren signals and location of siren signal source |
| CN212519401U (en) * | 2020-06-22 | 2021-02-09 | 深圳市鸿合创新信息技术有限责任公司 | Digital microphone circuit and pickup electronic equipment |
| CN111883161A (en) * | 2020-07-08 | 2020-11-03 | 东方通信股份有限公司 | Method and device for audio acquisition and position identification |
| CN112183647A (en) * | 2020-09-30 | 2021-01-05 | 国网山西省电力公司大同供电公司 | Transformer substation equipment sound fault detection and positioning method based on deep learning |
| CN112348052A (en) * | 2020-09-30 | 2021-02-09 | 国网山西省电力公司大同供电公司 | Power transmission and transformation equipment abnormal sound source positioning method based on improved EfficientNet |
| CN112710987A (en) * | 2020-12-28 | 2021-04-27 | 哈尔滨理工大学 | Microphone array system for identifying and positioning illegal whistle vehicle |
| CN112906795A (en) * | 2021-02-23 | 2021-06-04 | 江苏聆世科技有限公司 | Whistle vehicle judgment method based on convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 张全新: "《深度学习中的图像分类与对抗技术》", vol. 1, 北京理工大学出版社, pages: 11 - 13 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115061086A (en) * | 2022-05-12 | 2022-09-16 | 上海事凡物联网科技有限公司 | Moving target detection method based on micro-aperture microphone array |
| CN115169250A (en) * | 2022-09-05 | 2022-10-11 | 湖北工业大学 | Microphone array optimization method and system based on improved mixed integer differential evolution |
| CN115662436A (en) * | 2022-11-14 | 2023-01-31 | 北京探境科技有限公司 | Audio processing method and device, storage medium and intelligent glasses |
| CN117172135A (en) * | 2023-11-02 | 2023-12-05 | 山东省科霖检测有限公司 | Intelligent noise monitoring management method and system |
| CN117172135B (en) * | 2023-11-02 | 2024-02-06 | 山东省科霖检测有限公司 | Intelligent noise monitoring management method and system |
| CN117289208A (en) * | 2023-11-24 | 2023-12-26 | 北京瑞森新谱科技股份有限公司 | Sound source positioning method and device |
| CN117289208B (en) * | 2023-11-24 | 2024-02-20 | 北京瑞森新谱科技股份有限公司 | Sound source positioning method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114325214A (en) | Electric power online monitoring method based on microphone array sound source positioning technology | |
| JP7337953B2 (en) | Speech recognition method and device, neural network training method and device, and computer program | |
| CN110503969A (en) | A kind of audio data processing method, device and storage medium | |
| CN112885372B (en) | Intelligent diagnosis method, system, terminal and medium for power equipment fault sound | |
| CN107785029A (en) | Target voice detection method and device | |
| Yao et al. | A recursive denoising learning for gear fault diagnosis based on acoustic signal in real industrial noise condition | |
| CN110634476B (en) | Method and system for rapidly building robust acoustic model | |
| CN113077806B (en) | Audio processing method and device, model training method and device, medium and equipment | |
| CN111914705A (en) | Signal generation method and device for improving health state evaluation accuracy of reactor | |
| CN110322882A (en) | A kind of method and system generating mixing voice data | |
| CN111540342A (en) | Energy threshold adjusting method, device, equipment and medium | |
| Hao et al. | A Unified Framework for Low-Latency Speaker Extraction in Cocktail Party Environments. | |
| Ganjun et al. | Partial discharge pattern recognition of high voltage cables based on the stacked denoising autoencoder method | |
| Yuan et al. | Evolving multi-resolution pooling CNN for monaural singing voice separation | |
| CN113743301B (en) | Solid-state nanopore sequencing electric signal noise reduction processing method based on residual self-encoder convolutional neural network | |
| CN114563671A (en) | High-voltage cable partial discharge diagnosis method based on CNN-LSTM-Attention neural network | |
| Li et al. | Frame-Level Signal-to-Noise Ratio Estimation Using Deep Learning. | |
| Hou et al. | Multi-task learning for end-to-end noise-robust bandwidth extension | |
| CN111443328B (en) | Sound event detection and positioning method based on deep learning | |
| CN113095482A (en) | Raman amplifier gain value calculation method and system based on BP neural network | |
| CN112836788A (en) | Low-power-consumption deep learning network method for interference type identification | |
| CN114613384B (en) | Deep learning-based multi-input voice signal beam forming information complementation method | |
| CN110060699A (en) | A kind of single channel speech separating method based on the sparse expansion of depth | |
| CN116665681A (en) | Thunder identification method based on combined filtering | |
| CN115859055A (en) | Feature extraction method for multi-source heterogeneous big data in aircraft manufacturing process |
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 |