Transformer substation low-frequency sound source positioning method based on virtual microphone array and MUSIC algorithm

Abstract

The invention discloses a transformer substation low-frequency sound source positioning method based on a virtual microphone array and a MUSIC algorithm. The method comprises the steps that the expected position of a transformer substation low-frequency sound source is estimated; the method comprises the following steps of: performing interval sampling of position transformation in a plane directly facing an expected position of a low-frequency sound source by using a microphone array, and constructing a virtual microphone array; performing primary processing on the sampling signal to obtain a composite signal; and carrying out MUSIC algorithm processing on the synthesized signal to obtain a spatial map of transformer substation low-frequency sound source localization. Under the condition that the aperture of the microphone array is not changed, the large-aperture virtual microphone array is constructed, the physical aperture of the array is equivalently increased, the accuracy of low-frequency sound source positioning is improved, and the situation that the equipment size is increased and the manufacturing cost is increased due to the fact that the aperture of the microphone array is enlarged is avoided; the MUSIC algorithm is adopted to process microphone array data, the positioning resolution ratio is high, and meanwhile the method has the good resolution capacity for a plurality of noise sources in a short distance.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method for localizing low-frequency sound sources in a substation based on a virtual microphone array and a MUSIC algorithm. Background Art

[0002] The low-frequency noise generated by the numerous noise-generating devices in urban substations seriously disrupts the lives of surrounding residents, leading to a surge in demand for monitoring, locating, and protecting against low-frequency noise sources. In recent years, noise source localization technology has been widely used for noise localization in power transmission and transformation equipment. However, because the noise generated by power transmission and transformation equipment is primarily low-frequency, traditional microphone array sound source localization methods have low spatial resolution for low-frequency noise sources, making it difficult to detect multiple spatially close noise sources. Furthermore, the localization accuracy is also low, making it impossible to accurately identify the location of the noise source. Therefore, research on high-precision localization methods for low-frequency noise sources is of great significance for noise control, equipment optimization, and fault monitoring in substation equipment.

[0003] Because low-frequency noise has a long wavelength, using a traditional microphone array to locate its source requires increasing the array aperture to ensure accurate positioning. This increases the size of the equipment, increasing manufacturing costs. This increase in size also places additional restrictions on transportation and measurement site selection. To improve the accuracy of microphone array sound source localization systems in locating low-frequency noise sources without changing the original array aperture, a synthetic aperture method combined with a beamforming algorithm was used to study substation noise localization. The results show that the synthetic aperture method can improve the localization accuracy of low-frequency noise sources, but the spatial resolution of this method is not high enough, and it cannot effectively separate closely spaced sound sources.

[0004] Therefore, in order to further improve the positioning resolution of low-frequency noise sources in substations, it is urgent to develop a small-sized, high-resolution low-frequency sound source positioning method. Summary of the Invention

[0005] The present invention mainly aims to solve the problem of low resolution of existing substation low-frequency sound source localization technology. It provides a substation low-frequency sound source localization method based on a virtual microphone array and a MUSIC algorithm. Without changing the aperture of the microphone array, a large-aperture virtual microphone array is constructed, which equivalently increases the physical aperture of the array, improves the accuracy of the microphone array sound source localization method in locating low-frequency sound sources, and avoids the increase in equipment size and manufacturing cost caused by expanding the microphone array aperture. The MUSIC algorithm is used to process the microphone array data, so that the positioning resolution of the substation low-frequency noise is high, and the multiple noise sources at close range are better resolved.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solutions.

[0007] A method for localizing low-frequency sound sources in substations based on a virtual microphone array and a MUSIC algorithm includes the following steps: Step S1: estimating the expected location of the low-frequency sound source in the substation; Step S2: using the microphone array to perform interval sampling of changing positions in a plane directly opposite to the expected position of the low-frequency sound source to construct a virtual microphone array; Step S3: performing preliminary processing on the sampled signals of each sub-array in the virtual microphone array to obtain a synthesized signal; Step S4: Process the synthesized signal using the MUSIC algorithm to obtain a spatial spectrum for localizing the low-frequency sound source of the substation.

[0008] This paper proposes a high-resolution localization method for low-frequency sound sources in substations. Without changing the aperture of the microphone array, this method processes the sound signals (sampled signals) collected by the microphone array at different locations to synthesize a virtual large-aperture array, effectively increasing the array's physical aperture. The signal output values ​​of the virtual microphone array are used as initial values, and then processed using a high-resolution multiple signal classification (MUSIC) algorithm to locate and analyze the sound source. The MUSIC sound source localization method based on a virtual microphone array has high localization resolution for low-frequency sound sources and good resolution for multiple noise sources at close range. By synthesizing a large-aperture virtual microphone array, the method avoids increasing the microphone array aperture, which would increase the device size and manufacturing cost, thereby improving cost-effectiveness and applicability.

[0009] Preferably, in step S2, the microphone array performs interval sampling of changing positions in a plane directly opposite the expected position of the low-frequency sound source, and the position of the microphone array at the first sampling is defined as the reference position. During the interval sampling process, the distance l of the horizontal or vertical movement of the microphone array is greater than the diameter d of the microphone array. The position of the microphone array after each movement is a subarray of the virtual microphone array. When the microphone array moves n times in the same direction, the aperture D of the virtual microphone array is D = d × (n + 1). The present invention performs interval sampling of changing positions of the microphone array in a plane directly opposite the expected position of the sound source without changing the aperture of the microphone array, synthesizing a virtual large-aperture array, equivalently increasing the physical aperture of the array, avoiding the increase in equipment size and manufacturing cost caused by expanding the aperture of the microphone array, and improving economy and applicability.

[0010] Preferably, the specific process of step S3 includes the following steps: Step S31: Pass the sampled signal through a narrowband filter to obtain a filtered sampled signal; Step S32: transforming the filtered sampled signal using Hilbert transform to obtain an analytical form of the sampled signal; Step S33: Calculate the phase difference between the reference array and the subarrays, and synthesize the sampled signals from all subarrays to obtain a synthesized signal. This invention uses a virtual array method to synthesize the noise signals (sampled signals) collected by the microphone array at different locations into the noise signal of a large-aperture virtual microphone array. This avoids increasing the size of the equipment and manufacturing costs by enlarging the microphone array aperture to ensure positioning accuracy, and also avoids the increased size that would impose additional restrictions on transportation, measurement site selection, and other aspects.

[0011] Preferably, the specific process of step S4 includes the following steps: Step S41: Perform eigendecomposition on the array data covariance matrix corresponding to the synthetic signal to obtain mutually orthogonal signal subspace and noise subspace, which can be expressed as follows: Step S42: Using the orthogonality relationship between the signal subspace and the noise subspace, a spatial spectrum peak is constructed, and the minimum optimization search method is used to perform azimuth estimation by searching the spectrum peak. The formula is expressed as: Therefore, the spatial spectrum search formula for the synthetic signal y(t) using the MUSIC algorithm is expressed as: The MUSIC algorithm (Multiple Signal Classification algorithm) has higher spatial resolution and can distinguish two sound sources that are close to each other in space compared to traditional beamforming algorithms. It is especially suitable for locating low-frequency noise sources.

[0012] The MUSIC algorithm is a super-resolution DOA estimation algorithm based on the narrowband signal source hypothesis. The basic idea of ​​the MUSIC algorithm is to obtain mutually orthogonal signal subspace and noise subspace by eigendecomposing the array data covariance matrix. The orthogonality between the signal subspace and the noise subspace is used to construct the spatial spectrum peak. The signal direction estimation is achieved by searching for the spectrum peak. The details are as follows: The characteristic decomposition of the covariance matrix of array data is: Perform eigendecomposition on the covariance matrix, calculate the rank and eigenvalue of the output matrix, and construct a diagonal matrix consisting of the eigenvectors and eigenvalues ​​of the covariance matrix from the eigenvalues. D ,λ D+1 , ..., λ M ] and U=[u1,...,u D ,uD+1 ,...,u M ] are R YY The eigenvalue matrix and the eigenvector matrix corresponding to the eigenvalue. S and U N are the signal subspace and the noise subspace. Under ideal conditions, the signal space vector and the noise space vector are orthogonal: a H (θ)U N =0 In actual situations, due to the presence of noise, a H (θ) and U N It is not completely orthogonal and needs to be achieved through minimization search, where a(θ) is the direction vector: τ mθ is the time delay to the mth sensor in the spatial search direction θ. Considering the finite length of the actual received data matrix, the maximum likelihood estimate of the data covariance matrix is: Where: L represents the number of snapshots, the covariance matrix of the signal received by each microphone is calculated, and then the average value is calculated to obtain the maximum likelihood estimate. And the covariance matrix of the actual output is used Perform eigendecomposition to obtain the signal subspace part in the actual environment and the noise subspace part Due to the influence of the actual environment, a H (θ) and U N They are not completely orthogonal. To obtain accurate estimation results, the minimum optimization search is used to estimate the orientation: Since the array steering vector is orthogonal to the noise subspace, an extreme value will appear in the direction of signal incidence, and a peak will appear on the power spectrum image. The incident angle is calculated based on the coordinates corresponding to the peak position. The MUSIC spatial spectrum search formula is:

[0013] Preferably, in step S2, after the microphone array moves n times in the same direction, the aperture of the virtual microphone array is d(n+1). By synthesizing a large-aperture virtual microphone array, the present invention avoids increasing the size of the device and manufacturing costs caused by expanding the aperture of the microphone array, thereby improving cost-effectiveness and applicability.

[0014] As a preference, a spatial rectangular coordinate system is established with the center of the reference array at the reference position as the coordinate origin. Assuming that there are E point sound sources in the expected sound source area, the sampling signal p received by the array element at the coordinate origin isk for: 1≤k≤m Where B is the amplitude of the sampling signal; r k is the coordinate vector of the k-th point sound source; ω is the angular frequency of the sampling signal; is the wave number; c is the speed of sound in air.

[0015] As a preference, the sub-arrays at different positions after movement are synthesized to form a large-aperture virtual microphone array. In the large-aperture virtual microphone array, the sampling signal x received by the m-th element of the n-th sub-array is n，m (t) is: Where n is the position number of the subarray in the virtual microphone array; r n，m is the coordinate vector of the mth element of the nth subarray; θ n is the initial reference phase of the nth subarray, and 1≤n≤n x , 1≤m≤M.

[0016] Preferably, in step S33, the initial phases of the sampling signals received by the microphone arrays at different positions are different, and the initial reference phase difference Δθ between the nth subarray and the reference array is calculated using the sampling signal received by the reference array element at the reference position. n，0 , the formula is: Among them, θ n，m ,θ 0，m represent the initial reference phase of the mth array element in the nth subarray and the reference array respectively; x n，m 、x 0，m They represent the signals sampled by the m-th element in the n-th sub-array and the reference array respectively; arg represents the argument; and E represents the expectation.

[0017] Preferably, in step S33, the sampling signal received by the mth element of the nth subarray of the large aperture virtual microphone array at time t is x n，m (t), the sampled signal received by the mth element of the nth subarray in the synthesized large-aperture virtual microphone array is x n，m (t)′, where After the conversion of the above formula, the sampling signals received by all sub-arrays can be synchronized with the sampling signals received by the reference array, thereby converting the sampling signals received by sub-arrays at different positions into the overall receiving signal of the large-aperture virtual microphone array.

[0018] Preferably, in step S33, the synthesized signal y(t) is expressed as: Express y(t) in vector form: y(t)=A E s(t)+n(t) Among them, A E represents the direction matrix; s(t) represents the spatial signal vector; n(t) represents the noise vector.

[0019] Therefore, the advantages of the present invention are: (1) Without changing the aperture of the microphone array, a large-aperture virtual microphone array is constructed, which is equivalent to increasing the physical aperture of the array, thereby improving the accuracy of the microphone array sound source localization method in locating low-frequency sound sources, avoiding the increase in equipment size and manufacturing cost caused by expanding the microphone array aperture, and improving economy and applicability; (2) The MUSIC algorithm is used to process the microphone array data, which enables high positioning resolution for low-frequency noise in substations and good resolution for multiple noise sources at close range. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The present invention is a flowchart of a method for localizing a low-frequency sound source in a substation based on a virtual microphone array and a MUSIC algorithm.

[0021] Figure 2 4 is a flowchart of the MUSIC algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0023] Example 1.

[0024] A method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm, such as Figure 1 As shown, the following steps are included: Step S1: estimating the expected location of the low-frequency sound source in the substation, i.e. determining the approximate location of the low-frequency sound source to be measured; Step S2: A fixed reference microphone is placed between the measurement position and the array. The microphone continuously collects noise signals during the signal acquisition process. The microphone array is used to perform interval sampling at varying positions in a plane directly opposite the expected position of the low-frequency sound source to construct a virtual microphone array. Step S3: Preliminary processing is performed on the sampled signals of each subarray in the virtual microphone array to obtain a composite signal. After interval sampling at the moving position, the collected signal is passed through a narrowband filter and transformed using the Hilbert transform to obtain the analytical form of the signal. After calculating the phase difference between the reference array and the subarray, the sampled signals of all subarrays are synthesized. Step S4: The synthesized signal is processed by the MUSIC algorithm to obtain a spatial map for locating low-frequency sound sources in the substation. The synthesized signal is processed by the noise source localization algorithm, and the result obtained is the noise source localization result of the virtual array.

[0025] This embodiment proposes a high-resolution localization method for low-frequency sound sources in substations. Without changing the aperture of the microphone array, this method processes the sound signals (sampled signals) collected by the microphone array at different locations to synthesize a virtual large-aperture array, effectively increasing the physical aperture of the array. The signal output values ​​of the virtual microphone array are used as initial values, and the signal output values ​​are then processed using a high-resolution multiple signal classification (MUSIC) algorithm to locate and analyze the sound source. The MUSIC sound source localization method based on a virtual microphone array has high localization resolution for low-frequency sound sources and good resolution for multiple noise sources at close range. By synthesizing a large-aperture virtual microphone array, the method avoids increasing the size of the equipment and manufacturing costs caused by expanding the microphone array aperture, thereby improving cost-effectiveness and applicability.

[0026] In step S2, the microphone array performs interval sampling of varying positions within a plane directly opposite the expected location of the low-frequency sound source. The position of the microphone array at the first sampling is defined as the reference position. During the interval sampling process, the microphone array moves horizontally or vertically by a distance l greater than the diameter d of the microphone array. The position of the microphone array after each movement becomes a subarray of the virtual microphone array. After the microphone array moves n times in the same direction, the aperture D of the virtual microphone array is equal to d×(n+1). This embodiment performs interval sampling of varying positions of the microphone array within a plane directly opposite the expected location of the sound source without changing the aperture of the microphone array to synthesize a virtual large-aperture array, effectively increasing the physical aperture of the array. This avoids increasing the size of the device and manufacturing costs caused by expanding the aperture of the microphone array, thereby improving cost-effectiveness and applicability.

[0027] The specific process of step S3 is as follows: Figure 2 As shown, the following steps are included: Step S31: Pass the sampled signal through a narrowband filter to obtain a filtered sampled signal; Step S32: transforming the filtered sampled signal using Hilbert transform to obtain an analytical form of the sampled signal; Step S33: Calculate the phase difference between the reference array and the subarrays, and synthesize the sampled signals from all subarrays to obtain a synthesized signal. This embodiment uses a virtual array method to synthesize the noise signals (sampled signals) collected at different locations of the microphone array into the noise signal of a large-aperture virtual microphone array. This avoids increasing the size of the equipment and manufacturing costs by enlarging the microphone array aperture to ensure positioning accuracy, and also avoids the additional restrictions imposed by the increased size on transportation, measurement site selection, and other aspects.

[0028] The specific process of step S4 includes the following steps: Step S41: Perform eigendecomposition on the array data covariance matrix corresponding to the synthetic signal to obtain mutually orthogonal signal subspace and noise subspace, which can be expressed as follows: Step S42: Using the orthogonality relationship between the signal subspace and the noise subspace, a spatial spectrum peak is constructed, and the minimum optimization search method is used to perform azimuth estimation by searching the spectrum peak. The formula is expressed as: Therefore, the spatial spectrum search formula for the synthetic signal y(t) using the MUSIC algorithm is expressed as: The MUSIC algorithm (Multiple Signal Classification algorithm) has higher spatial resolution and can distinguish two sound sources that are close to each other in space compared to traditional beamforming algorithms. It is especially suitable for locating low-frequency noise sources.

[0029] The MUSIC algorithm is a super-resolution DOA estimation algorithm based on the narrowband signal source hypothesis. The basic idea of ​​the MUSIC algorithm is to obtain mutually orthogonal signal subspace and noise subspace by eigendecomposing the array data covariance matrix. The orthogonality between the signal subspace and the noise subspace is used to construct the spatial spectrum peak. The signal direction estimation is achieved by searching for the spectrum peak. The details are as follows: The characteristic decomposition of the covariance matrix of array data is: Perform eigendecomposition on the covariance matrix, calculate the rank and eigenvalue of the output matrix, and construct a diagonal matrix consisting of the eigenvectors and eigenvalues ​​of the covariance matrix from the eigenvalues. D ,λ D+1 , ..., λ M ] and U=[u1,...,u D ,u D+1 ,...,u M ] are R YYThe eigenvalue matrix and the eigenvector matrix corresponding to the eigenvalue. S and U N are the signal subspace and the noise subspace. Under ideal conditions, the signal space vector and the noise space vector are orthogonal: a H (θ)U N =0 In actual situations, due to the presence of noise, a H (θ) and U N It is not completely orthogonal and needs to be achieved through minimization search, where a(θ) is the direction vector: τ mθ is the time delay to the mth sensor in the spatial search direction θ. Considering the finite length of the actual received data matrix, the maximum likelihood estimate of the data covariance matrix is: Where: L represents the number of snapshots, the covariance matrix of the signal received by each microphone is calculated, and then the average value is calculated to obtain the maximum likelihood estimate. And the covariance matrix of the actual output is used Perform eigendecomposition to obtain the signal subspace part in the actual environment and the noise subspace part Due to the influence of the actual environment, a H (θ) and U N They are not completely orthogonal. To obtain accurate estimation results, the minimum optimization search is used to estimate the orientation: Since the array steering vector is orthogonal to the noise subspace, an extreme value will appear in the direction of signal incidence, and a peak will appear on the power spectrum image. The incident angle is calculated based on the coordinates corresponding to the peak position. The MUSIC spatial spectrum search formula is:

[0030] In step S2, after the microphone array moves n times in the same direction, the aperture of the virtual microphone array is d(n+1). This embodiment synthesizes a large-aperture virtual microphone array to avoid increasing the size of the device and manufacturing costs caused by expanding the microphone array aperture, thereby improving cost-effectiveness and applicability.

[0031] A spatial rectangular coordinate system is established with the center of the reference array at the reference position as the coordinate origin. Assuming that there are E point sound sources in the expected sound source area, the sampling signal p received by the array element at the coordinate origin is k for: 1≤k≤m Where B is the amplitude of the sampling signal; r k is the coordinate vector of the k-th point sound source; ω is the angular frequency of the sampling signal; is the wave number; c is the speed of sound in air.

[0032] After the movement, the sub-arrays at different positions are synthesized to form a large-aperture virtual microphone array. In the large-aperture virtual microphone array, the sampling signal x received by the m-th element of the n-th sub-array is n，m (t) is: Where n is the position number of the subarray in the virtual microphone array; r n，m is the coordinate vector of the mth element of the nth subarray; θ n is the initial reference phase of the nth subarray, and 1≤n≤n x , 1≤m≤M.

[0033] In step S33, the initial phases of the sampling signals received by the microphone arrays at different positions are different. The initial reference phase difference Δθ between the nth subarray and the reference array is calculated using the sampling signals received by the reference array element at the reference position. n，0 , the formula is: Among them, θ n，m ,θ 0，m represent the initial reference phase of the mth array element in the nth subarray and the reference array respectively; x n，m 、x 0，m They represent the signals sampled by the m-th element in the n-th sub-array and the reference array respectively; arg represents the argument; and E represents the expectation.

[0034] In step S33, the sampling signal received by the mth element of the nth subarray of the large aperture virtual microphone array at time t is x n，m (t), the sampled signal received by the mth element of the nth subarray in the synthesized large-aperture virtual microphone array is x n，m (t)′, where After the conversion of the above formula, the sampling signals received by all sub-arrays can be synchronized with the sampling signals received by the reference array, thereby converting the sampling signals received by sub-arrays at different positions into the overall receiving signal of the large-aperture virtual microphone array.

[0035] In step S33, if the qth microphone of the virtual large aperture array corresponds to the mth element of the nth subarray, then y q (t) = x n，m(t)′, assuming that there are E narrowband signal sources in the space, which are incident on the virtual array composed of three arrays, where y(t) is the signal source. The received signal of the entire virtual array can be expressed as the composite signal y(t) expressed as: Express y(t) in vector form: y(t)=A E s(t)+n(t) Among them, y(t) is the data of the signal received by the array, with a dimension of 3m×1, n(t) is the noise vector with a dimension of 3m×1 received by the array, s(t) is the spatial signal vector with a dimension of E×1, A E Represents the array direction matrix with dimension 3m×E.

[0036] Example 2.

[0037] A method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm, such as Figure 1 As shown, the following steps are included: Step S1: estimating the expected location of the low-frequency sound source in the substation, i.e. determining the approximate location of the low-frequency sound source to be measured; Step S2: A fixed reference microphone is placed between the measurement position and the array. The microphone continuously collects noise signals during the signal acquisition process. The microphone array is used to perform interval sampling at varying positions in a plane directly opposite the expected position of the low-frequency sound source to construct a virtual microphone array. Step S3: Preliminary processing is performed on the sampled signals of each subarray in the virtual microphone array to obtain a composite signal. After interval sampling at the moving position, the collected signal is passed through a narrowband filter and transformed using the Hilbert transform to obtain the analytical form of the signal. After calculating the phase difference between the reference array and the subarray, the sampled signals of all subarrays are synthesized. Step S4: The synthesized signal is processed by the MUSIC algorithm to obtain a spatial map for locating low-frequency sound sources in the substation. The synthesized signal is processed by the noise source localization algorithm, and the result obtained is the noise source localization result of the virtual array.

[0038] In step S2, the microphone array performs interval sampling of varying positions within a plane directly opposite the expected location of the low-frequency sound source. The position of the microphone array at the first sampling is defined as the reference position. During the interval sampling process, the microphone array moves horizontally or vertically by a distance l greater than the diameter d of the microphone array. The position of the microphone array after each movement becomes a subarray of the virtual microphone array. After the microphone array moves n times in the same direction, the aperture D of the virtual microphone array is equal to d×(n+1). This embodiment performs interval sampling of varying positions of the microphone array within a plane directly opposite the expected location of the sound source without changing the aperture of the microphone array to synthesize a virtual large-aperture array, effectively increasing the physical aperture of the array. This avoids increasing the size of the device and manufacturing costs caused by expanding the aperture of the microphone array, thereby improving cost-effectiveness and applicability.

[0039] A spatial rectangular coordinate system is established with the center of the reference array at the reference position as the coordinate origin. Assuming that there are E point sound sources in the expected sound source area, the sampling signal p received by the array element at the coordinate origin is k for: 1≤k≤m Where B is the amplitude of the sampling signal; r k is the coordinate vector of the k-th point sound source; ω is the angular frequency of the sampling signal; is the wave number; c is the speed of sound in air.

[0040] After the movement, the sub-arrays at different positions are synthesized to form a large-aperture virtual microphone array. In the large-aperture virtual microphone array, the sampling signal x received by the m-th element of the n-th sub-array is n，m (t) is: Where n is the position number of the subarray in the virtual microphone array; r n，m is the coordinate vector of the mth element of the nth subarray; θ n is the initial reference phase of the nth subarray, and 1≤n≤n x , 1≤m≤M.

[0041] The specific process of step S3 is as follows: Figure 2 As shown, the following steps are included: Step S31: Pass the sampled signal through a narrowband filter to obtain a filtered sampled signal; Step S32: transforming the filtered sampled signal using Hilbert transform to obtain an analytical form of the sampled signal; Step S33: Calculate the phase difference between the reference array and the subarrays, and synthesize the sampled signals from all subarrays to obtain a synthesized signal. This embodiment uses a virtual array method to synthesize the noise signals (sampled signals) collected at different locations of the microphone array into the noise signal of a large-aperture virtual microphone array. This avoids increasing the size of the equipment and manufacturing costs by enlarging the microphone array aperture to ensure positioning accuracy, and also avoids the additional restrictions imposed by the increased size on transportation, measurement site selection, and other aspects.

[0042] In step S33, the initial phases of the sampling signals received by the microphone arrays at different positions are different. The initial reference phase difference Δθ between the nth subarray and the reference array is calculated using the sampling signals received by the reference array element at the reference position. n，0 , the formula is: Among them, θ n，m ,θ 0，m represent the initial reference phase of the mth array element in the nth subarray and the reference array respectively; x n，m 、x 0，m They represent the signals sampled by the m-th element in the n-th sub-array and the reference array respectively; arg represents the argument; and E represents the expectation.

[0043] In step S33, the sampling signal received by the mth element of the nth subarray of the large aperture virtual microphone array at time t is x n，m (t), the sampled signal received by the mth element of the nth subarray in the synthesized large-aperture virtual microphone array is x n，m (t)′, where After the conversion of the above formula, the sampling signals received by all sub-arrays can be synchronized with the sampling signals received by the reference array, thereby converting the sampling signals received by sub-arrays at different positions into the overall receiving signal of the large-aperture virtual microphone array.

[0044] In step S33, if the qth microphone of the virtual large aperture array corresponds to the mth element of the nth subarray, then y q (t) = x n，m (t)′, assuming that there are E narrowband signal sources in the space, which are incident on the virtual array composed of three arrays, where y(t) is the signal source. The received signal of the entire virtual array can be expressed as the composite signal y(t) expressed as: Express y(t) in vector form: y(t)=A E s(t)+n(t) Among them, y(t) is the data of the signal received by the array, with a dimension of 3m×1, n(t) is the noise vector with a dimension of 3m×1 received by the array, s(t) is the spatial signal vector with a dimension of E×1, A E Represents the array direction matrix with dimension 3m×E.

[0045] The specific process of step S4 includes the following steps: Step S41: Perform eigendecomposition on the array data covariance matrix corresponding to the synthetic signal to obtain mutually orthogonal signal subspace and noise subspace, which can be expressed as follows: Step S42: Using the orthogonality relationship between the signal subspace and the noise subspace, a spatial spectrum peak is constructed, and the minimum optimization search method is used to perform azimuth estimation by searching the spectrum peak. The formula is expressed as: Therefore, the spatial spectrum search formula for the synthetic signal y(t) using the MUSIC algorithm is expressed as: The MUSIC algorithm (Multiple Signal Classification algorithm) has higher spatial resolution and can distinguish two sound sources that are close to each other in space compared to traditional beamforming algorithms. It is especially suitable for locating low-frequency noise sources.

[0046] The MUSIC algorithm is a super-resolution DOA estimation algorithm based on the narrowband signal source hypothesis. The basic idea of ​​the MUSIC algorithm is to obtain mutually orthogonal signal subspace and noise subspace by eigendecomposing the array data covariance matrix. The orthogonality between the signal subspace and the noise subspace is used to construct the spatial spectrum peak. The signal direction estimation is achieved by searching for the spectrum peak. The details are as follows: The characteristic decomposition of the covariance matrix of array data is: Perform eigendecomposition on the covariance matrix, calculate the rank and eigenvalue of the output matrix, and construct a diagonal matrix consisting of the eigenvectors and eigenvalues ​​of the covariance matrix from the eigenvalues. D ,λ D+1 , ..., λ M ] and U=[u1,...,u D ,u D+1 ,.....,u M ] are R YY The eigenvalue matrix and the eigenvector matrix corresponding to the eigenvalue. S and U Nare the signal subspace and the noise subspace. Under ideal conditions, the signal space vector and the noise space vector are orthogonal: a H (θ)U N =0 In actual situations, due to the presence of noise, a H (θ) and U N It is not completely orthogonal and needs to be achieved through minimization search, where a(θ) is the direction vector: τ mθ is the time delay to the mth sensor in the spatial search direction θ. Considering the finite length of the actual received data matrix, the maximum likelihood estimate of the data covariance matrix is: Where: L represents the number of snapshots, the covariance matrix of the signal received by each microphone is calculated, and then the average value is calculated to obtain the maximum likelihood estimate. And the covariance matrix of the actual output is used Perform eigendecomposition to obtain the signal subspace part in the actual environment and the noise subspace part Due to the influence of the actual environment, a H (θ) and U N They are not completely orthogonal. To obtain accurate estimation results, the minimum optimization search is used to estimate the orientation: Since the array steering vector is orthogonal to the noise subspace, an extreme value will appear in the direction of signal incidence, and a peak will appear on the power spectrum image. The incident angle is calculated based on the coordinates corresponding to the peak position. The MUSIC spatial spectrum search formula is:

[0047] Although the traditional MUSIC algorithm has a high spatial resolution, its resolution is still related to the frequency of the noise source. As the frequency of the noise source decreases, the spatial resolution also decreases. In order to further improve the resolution of low-frequency noise sources, the present invention proposes a MUSIC algorithm based on a virtual array. By constructing a virtual array, the array aperture is increased without increasing the physical diameter of the array, thereby improving the spatial resolution of the array. The virtual array moves the array to multiple positions in the same plane, and then through certain phase and spatial position compensation, the different positions of the array are virtualized into a large-aperture array to locate the noise source and obtain the positioning effect of the large-aperture array. After determining the approximate position of the noise source to be measured, a fixed reference microphone is placed between the measurement position and the array. This microphone continuously collects noise signals during the signal acquisition process. The microphone array performs interval sampling of changing positions in the plane opposite to the noise source position. The position of the microphone array at the first sampling is defined as the reference position. To ensure that there is no overlap in the virtual array elements, the microphone array moves horizontally or vertically, a distance l greater than the array diameter d, during the interval sampling process. The position of the array after each movement becomes a subarray of the virtual array. After the array moves n times in the same direction, the aperture of the virtual array is synthesized as D = d × (n + 1), where D is the aperture of the virtual array. After the interval sampling of the moving position, the collected signal is passed through a narrowband filter and transformed using the Hilbert transform to obtain the signal's analytical form. After calculating the phase difference between the reference array and the subarray, the sampled signals of all subarrays are synthesized and processed by the noise source localization algorithm. The result is the noise source localization result of the virtual array.

[0048] This embodiment proposes a high-resolution localization method for low-frequency sound sources in substations. Without changing the aperture of the microphone array, this method processes the sound signals (sampled signals) collected by the microphone array at different locations to synthesize a virtual large-aperture array, effectively increasing the physical aperture of the array. The signal output values ​​of the virtual microphone array are used as initial values, and the signal output values ​​are then processed using a high-resolution multiple signal classification (MUSIC) algorithm to locate and analyze the sound source. The MUSIC sound source localization method based on a virtual microphone array has high localization resolution for low-frequency sound sources and good resolution for multiple noise sources at close range. By synthesizing a large-aperture virtual microphone array, the method avoids increasing the size of the equipment and manufacturing costs caused by expanding the microphone array aperture, thereby improving cost-effectiveness and applicability.

[0049] The above content is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm, characterized in that: The following steps are involved: Step S1: estimating the expected location of the low-frequency sound source in the substation; Step S2: using the microphone array to perform interval sampling of changing positions in a plane directly opposite to the expected position of the low-frequency sound source to construct a virtual microphone array; Step S3: performing preliminary processing on the sampled signals of each sub-array in the virtual microphone array to obtain a synthesized signal; Step S4: Process the synthesized signal using the MUSIC algorithm to obtain a spatial spectrum for localizing the low-frequency sound source of the substation.

2. The method for localizing low-frequency sound sources in substations based on a virtual microphone array and a MUSIC algorithm according to claim 1, wherein: In step S2, the microphone array performs interval sampling of the changing position in the plane directly opposite the expected position of the low-frequency sound source. The position of the microphone array at the first sampling is defined as the reference position. During the interval sampling process, the horizontal or vertical movement distance l of the microphone array is greater than the diameter d of the microphone array. The position of the microphone array after each movement is a subarray of the virtual microphone array.

3. The method for localizing low-frequency sound sources in substations based on a virtual microphone array and a MUSIC algorithm according to claim 1 or 2, characterized in that: The specific process of step S3 includes the following steps: Step S31: Pass the sampled signal through a narrowband filter to obtain a filtered sampled signal; Step S32: transforming the filtered sampled signal using Hilbert transform to obtain an analytical form of the sampled signal; Step S33: Calculate the phase difference between the reference array and the sub-array, synthesize the sampled signals of all sub-arrays, and obtain a synthesized signal.

4. The method for localizing low-frequency sound sources in substations based on a virtual microphone array and a MUSIC algorithm according to claim 1 or 2, characterized in that: The specific process of step S4 includes the following steps: Step S41: performing eigendecomposition on the array data covariance matrix corresponding to the synthetic signal to obtain a mutually orthogonal signal subspace and noise subspace; Step S42: constructing a spatial spectrum peak by using the orthogonality relationship between the signal subspace and the noise subspace, and performing azimuth estimation by searching for the spectrum peak using a minimum optimization search method.

5. The method for localizing low-frequency sound sources in substations based on a virtual microphone array and a MUSIC algorithm according to claim 2, wherein: In step S2, after the microphone array moves n times in the same direction, the aperture of the virtual microphone array is d(n+1).

6. The method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm according to claim 5, characterized in that: A spatial rectangular coordinate system is established with the center of the reference array at the reference position as the coordinate origin. Assuming that there are E point sound sources in the expected sound source area, the sampling signal received by the array element at the coordinate origin is Where B is the amplitude of the sampling signal; r k is the coordinate vector of the k-th point sound source; ω is the angular frequency of the sampling signal; is the wave number, 1≤k≤m; c is the speed of sound in air.

7. The method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm according to claim 6, characterized in that: After moving, the sub-arrays at different positions are synthesized to form a large-aperture virtual microphone array. In the large-aperture virtual microphone array, the sampling signal received by the m-th element of the n-th sub-array is Where n is the position number of the subarray in the virtual microphone array; r n,m is the coordinate vector of the mth element of the nth subarray; θ n is the initial reference phase of the nth subarray, and 1≤n≤n x ,1≤m≤M.

8. The method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm according to claim 3, characterized in that: In step S33, the initial reference phase difference between the nth subarray and the reference array is calculated using the sampling signal received by the reference array element at the reference position. Among them, θ n,m ,θ 0,m represent the initial reference phase of the mth array element in the nth subarray and the reference array respectively; x n,m 、x 0,m They represent the signals sampled by the m-th element in the n-th sub-array and the reference array respectively; arg represents the argument; and E represents the expectation.

9. The method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm according to claim 8, characterized in that: In step S33, the sampling signal received by the mth element of the nth subarray of the large aperture virtual microphone array at time t is x n,m (t), the sampled signal received by the mth element of the nth subarray in the synthesized large-aperture virtual microphone array is 10. The method for localizing low-frequency sound sources in substations based on virtual microphone array and MUSIC algorithm according to claim 9, characterized in that: In step S33, the synthesized signal y(t) is expressed in vector form: y(t)=A E s(t)+n(t),A E represents the direction matrix; s(t) represents the spatial signal vector; n(t) represents the noise vector.

Images