A method for locating birds on a power transmission line based on sound source positioning

Abstract

The application provides a power transmission line bird position positioning method based on sound source positioning, and relates to the technical field of power transmission line online monitoring.The application collects multi-channel voiceprint data through uniformly arranging a circular microphone array on a power transmission tower, adaptively determines the number of layers of variational mode decomposition by using spectrum entropy and kurtosis analysis, and filters effective modes to reconstruct a high signal-to-noise ratio bird call signal; then, a two-dimensional multiple signal classification spatial spectrum is constructed by using characteristic decomposition of a covariance matrix, and a spectrum peak is extracted in a search domain of an azimuth angle and a pitch angle to determine the position of the bird sound source.The application realizes all-weather, high-precision and automatic monitoring and rapid early warning of bird damage of the power transmission line.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for power transmission lines, and in particular to a method for locating birds on power transmission lines based on sound source localization. Background Technology

[0002] Transmission lines are exposed to the wild environment for a long time. Bird activities (such as nesting, defecating, and roosting) are key factors that cause major accidents such as insulator flashover and line short circuits, which seriously threaten the safe and stable operation of the power grid.

[0003] Current mainstream bird monitoring methods mainly rely on manual inspections or video surveillance based on visible light / infrared cameras. However, these methods face significant bottlenecks in practical applications. First, they have poor environmental adaptability. In adverse weather conditions such as fog, rain, and nighttime, the performance of optical equipment deteriorates sharply or even fails, making effective monitoring impossible around the clock. Second, their positioning accuracy is insufficient. Birds have a wide and random range of activity, and traditional acoustic positioning methods are easily affected by complex background noise such as wind noise, mechanical vibrations of power transmission lines, and environmental noise, leading to large deviations in positioning results and making it difficult to accurately pinpoint the location of birds. Finally, they lack real-time capability. The massive amounts of acoustic data collected often require manual post-processing, which is inefficient and cannot provide effective support for timely early warning and rapid response.

[0004] Therefore, existing technologies, especially acoustic positioning schemes such as single microphone arrays, are difficult to effectively separate bird voiceprints from interference signals in the complex noise environment of power transmission lines. There is an urgent need to develop a new bird positioning method with high noise resistance, high positioning accuracy and real-time response capability to achieve proactive, accurate and efficient control of bird activity around power transmission lines. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method for locating birds on power transmission lines based on sound source localization. By combining adaptive acoustic enhancement with two-dimensional sound source localization, it achieves all-weather, high-precision, automated monitoring and rapid early warning of bird locations on power transmission lines.

[0006] To achieve the above objectives, the present invention provides the following solution: A method for locating birds on power transmission lines based on sound source localization, comprising: A uniform circular microphone array is deployed on the transmission tower to collect multi-channel raw acoustic data around the line, forming a noisy bird call signal. The noisy bird call signal is subjected to spectral analysis to calculate spectral entropy and kurtosis. Based on the criterion that the ratio of reconstructed energy to original energy tends to be stable, the number of decomposition levels of variational mode decomposition is adaptively determined and the variational mode decomposition parameters are initialized. The noisy bird call signal is decomposed into multiple intrinsic mode functions using a variational mode decomposition algorithm. The kurtosis or correlation of each intrinsic mode function is calculated to filter out effective modes containing bird calls. The effective modes are then superimposed and reconstructed to obtain a high signal-to-noise ratio bird call signal. The high signal-to-noise ratio bird call signal is constructed into a microphone array received signal matrix. The covariance matrix of the microphone array received signal matrix is ​​calculated and eigenvalue decomposition is performed to obtain the signal subspace and noise subspace. Based on the noise subspace and array manifold vector, a two-dimensional multi-signal classification spatial spectrum is constructed. Spectral peaks are searched in the two-dimensional search domain of azimuth and elevation angles, and the azimuth and elevation angles of the bird sound source corresponding to the spectral peaks are output to locate the bird's position.

[0007] Preferably, the uniform circular microphone array satisfies the half-wavelength sampling condition of the wavelength corresponding to the highest frequency; the spacing between adjacent microphones in the uniform circular microphone array is less than half of the wavelength corresponding to the highest frequency of the monitored bird calls, and the number of sensors in the uniform circular microphone array is greater than the number of bird sound sources to be located.

[0008] Preferably, the spectrum analysis includes: Perform a Fast Fourier Transform on the noisy bird call signal to obtain the dominant frequency distribution; The spectral entropy and kurtosis are calculated based on the dominant frequency distribution to drive the adaptive determination of the number of decomposition layers.

[0009] Preferably, adaptively determining the number of decomposition levels for variational mode decomposition includes: Starting from a preset initial candidate decomposition level, the number of levels is incremented by 1 each time. Modal decomposition and reconstruction are performed separately, and the ratio of the reconstructed signal energy to the original signal energy corresponding to each candidate decomposition level is calculated. When the energy ratio first stabilizes as the number of candidate decomposition layers increases, the corresponding number of candidate decomposition layers is determined as the number of decomposition layers.

[0010] Preferably, the variational mode decomposition parameters include a penalty factor, a convergence threshold, a maximum number of iterations, and an initial center frequency; wherein the convergence threshold is 0.000001, the maximum number of iterations is 500, the initial center frequency is determined by the dominant frequency of the noisy bird call signal obtained by the fast Fourier transform, and the penalty factor is adaptively adjusted according to the signal bandwidth.

[0011] Preferably, filtering effective modalities containing bird calls includes: Calculate the kurtosis of each intrinsic mode function and arrange them in descending order of kurtosis, then accumulate the corresponding energy percentages in sequence. When the cumulative energy percentage reaches or exceeds 90% and the estimated signal-to-noise ratio is greater than 3 dB, the corresponding number of intrinsic mode functions are determined as the effective modes.

[0012] Preferably, the estimated signal-to-noise ratio is obtained by real-time estimation of the background noise energy corresponding to the residual intrinsic mode functions that were not selected as the effective mode, and is used to adaptively adjust the dynamic truncation position of the effective mode.

[0013] Preferably, the azimuth search range in the two-dimensional search domain is 0 degrees to 360 degrees, and the pitch search range is 0 degrees to 90 degrees. The global spectral peak is obtained as the azimuth and pitch output of the bird sound source by traversing the two-dimensional search domain.

[0014] Preferably, after outputting the azimuth and elevation angles of the bird sound source, the azimuth and elevation angles of the bird sound source, the corresponding positioning time, and the high signal-to-noise ratio bird call signal are uploaded to the edge computing terminal device on the power transmission tower to realize real-time detection, positioning, and early warning linkage of bird activities.

[0015] Preferably, the uniform circular microphone array uses omnidirectional microphones, and each omnidirectional microphone is equipped with a protective shell to ensure the long-term stable acquisition capability of the omnidirectional microphone in a wide temperature range and high humidity field environment.

[0016] The present invention discloses the following technical effects: This invention collects multi-channel bird calls by deploying a uniform circular microphone array on transmission towers and introduces an adaptive variational mode decomposition and mode screening reconstruction mechanism. This enables the high signal-to-noise ratio bird call signals to maintain the separability of bird voiceprints even under strong interference from wind noise, line mechanical vibration, and environmental noise. This fundamentally overcomes the shortcomings of traditional acoustic positioning, which relies heavily on preset decomposition parameters, is prone to insufficient or excessive decomposition, and has weak noise resistance, resulting in large positioning errors. This invention achieves stable voiceprint enhancement in complex outdoor scenarios.

[0017] This invention uses the high signal-to-noise ratio bird call signal as input to a two-dimensional multi-signal classification spatial spectrum estimation. It accurately divides the signal subspace and noise subspace through covariance matrix eigenvalue decomposition and obtains the global spectral peak in the two-dimensional search domain of azimuth and elevation angles. Thus, it simultaneously outputs the azimuth and elevation angles of the bird sound source, achieving high-precision positioning of the bird's spatial location. This overcomes the shortcomings of existing technologies that cannot accurately give the three-dimensional orientation of birds based solely on video or conventional acoustic methods under conditions of multiple noise, multiple paths, and non-line-of-sight.

[0018] This invention automates the closed-loop process of signal decomposition, modal reconstruction, spatial spectrum calculation, and spectral peak search. It also supports the real-time uploading of the azimuth and elevation angles of bird sound sources and their corresponding positioning times to edge computing terminal devices on power transmission towers. This enables all-weather, human-free online detection and early warning linkage. It significantly improves the speed of bird damage risk detection and the efficiency of handling, addressing the problems of existing visible light or infrared monitoring failing in foggy or rainy nights and the lack of real-time performance due to the need for manual post-processing of acoustic data. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the IVMD decomposition process provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a uniform circular array of M elements provided in an embodiment of the present invention; Figure 4 This is a structural diagram of a sensor array provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The purpose of this invention is to provide a method for locating birds on power transmission lines based on sound source localization. By utilizing multi-channel acoustic sensing combined with adaptive mode decomposition and two-dimensional spatial spectrum localization technology, the technical bottleneck of accurately obtaining the location of birds in the complex noise environment of power transmission lines is solved.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1As shown, this invention provides a method for locating birds on power transmission lines based on sound source localization, comprising: Step 100: Deploy a uniform circular microphone array on the transmission tower to collect multi-channel raw acoustic data around the line and form a noisy bird call signal; Step 200: Perform spectral analysis on the noisy bird call signal, calculate the spectral entropy and kurtosis, and adaptively determine the number of decomposition levels of variational mode decomposition and initialize the variational mode decomposition parameters based on the criterion that the ratio of reconstructed energy to original energy tends to be stable. Step 300: The noisy bird call signal is decomposed into multiple intrinsic mode functions using the variational mode decomposition algorithm. The kurtosis or correlation of each intrinsic mode function is calculated to filter out effective modes containing bird calls. The effective modes are then superimposed and reconstructed to obtain a high signal-to-noise ratio bird call signal. Step 400: Construct the high signal-to-noise ratio bird call signal into a microphone array received signal matrix, calculate the covariance matrix of the microphone array received signal matrix and perform eigenvalue decomposition to obtain the signal subspace and noise subspace; Step 500: Construct a two-dimensional multi-signal classification spatial spectrum based on the noise subspace and array manifold vector, search for spectral peaks in the two-dimensional search domain of azimuth and elevation angles, and output the azimuth and elevation angles of the bird sound source corresponding to the spectral peaks to locate the bird's position.

[0025] Specifically, this embodiment aims to provide an improved acoustic array bird localization method that deeply integrates the VMD algorithm and the two-dimensional multiple signal classification (MUSIC) algorithm, systematically solving the core problems existing in the current technology for bird monitoring of power transmission lines. The specific objectives include: First, significantly improving noise immunity. This is achieved by introducing signal energy confirmation to improve the VMD algorithm, and using the improved IVMD algorithm to adaptively decompose the collected bird acoustic signature signals into a series of intrinsic mode functions (IMFs). The kurtosis criterion is used as a key indicator to accurately select IMF components containing effective bird acoustic signature information, while effectively suppressing strong background noise interference such as wind noise and equipment vibration, thus enhancing the signal-to-noise ratio. Second, optimizing positioning accuracy. A uniform circular microphone array structure is adopted, combined with the two-dimensional MUSIC algorithm. The covariance matrix is ​​constructed using the clean signal reconstructed by IVMD, and eigenvalue decomposition is performed to accurately separate the signal subspace and noise subspace. Then, a noise matrix is ​​constructed, and a two-dimensional spatial spectrum is calculated. By searching for spectral peaks, high-precision positioning of the bird source in spatial azimuth and elevation angles is achieved, overcoming the positioning deviation of traditional methods in complex environments. Third, it enables real-time monitoring and early warning. The entire signal processing and positioning process is designed as an automated and modular algorithm that can be directly deployed on edge computing terminal devices on transmission towers. This enables real-time detection, positioning, and data uploading of bird activities, providing timely and accurate bird information for the power grid's active defense system, and supporting rapid decision-making and response.

[0026] Specifically, the variational mode decomposition (VMD) algorithm involved in this embodiment is conceptualized as follows: Variational mode decomposition (VMD) is an adaptive, non-recursive method for variational mode and signal processing. The key step is variational solution. By iteratively solving for the center frequency and bandwidth of each mode component, effective separation of the intrinsic mode functions (IMFs) in the frequency domain is achieved. The specific process is as follows: (1) Converting the spectrum of each mode to baseband yields: ; (2) Thus, the constrained variational model can be expressed as: ; (3) Introducing a penalty factor, the augmented Lagrange expression is: ; Through the , , The iterative update yields the optimal solution, which is the target IMF's { } and frequency { }

[0027] Next, kurtosis is used to filter the intrinsic mode functions (IMFs) of the signal as a reference for signal reconstruction. Kurtosis is a fourth-order statistic used to measure the sharpness of the signal distribution. For a discrete signal x, the expression for kurtosis is: ; In the formula: , This is expressed as the mean and standard deviation of the signal; It is the fourth-order mathematical expectation of the signal.

[0028] Since VMD decomposition is a variational model solution, the number of decomposition levels K needs to be determined in advance. This invention addresses this problem by proposing a method based on signal energy determination, the specific process of which is as follows: (1) The acquired noisy discharge signal can be expressed as: ; in, ( t ) is the reconstructed signal, n ( t () represents noise signal.

[0029] (2) Let ER be the ratio of the energy of the reconstructed signal to that of the original signal. The expression for ER is shown in equation (6): ; (3) As the number of decomposition layers increases, ER will tend to stabilize, and the first stable K value is the optimal K value. The specific decomposition process is as follows: Figure 2 As shown.

[0030] Next, we will introduce signal reconstruction methods. Kurtosis is a fourth-order statistic used to measure the sharpness of a signal distribution. For a discrete signal x, the expression for kurtosis is: ; In the formula: , This is expressed as the mean and standard deviation of the signal; It is the fourth-order mathematical expectation of the signal.

[0031] Conventional methods use a value greater than 3 as a screening criterion. However, in noisy environments, this method struggles to accurately screen IMFs, leading to a decrease in denoising effectiveness. Therefore, this embodiment proposes an Energy Kurtosis Indicator (EKI) as an evaluation standard.

[0032] The kurtosis of each IMF is calculated and sorted in descending order of kurtosis. Then the energy of each IMF is calculated, and the signal, EKI, is reconstructed based on the estimated signal-to-noise ratio.

[0033] The estimated signal-to-noise ratio expression is as follows: ; Where, h ( t ) is the estimated discharge signal, n ( t ) is the noise signal.

[0034] Suppose the noisy discharge signal is decomposed into K IMFs, calculate the kurtosis value of each IMF and arrange them in descending order: ; Calculate the energy of each IMF, corresponding to the kurtosis value, denoted as E i (0 < i K), and the energy of the noisy signal is denoted as E0. When i exactly satisfies the following formula, the first i IMFs are used to reconstruct the signal.

[0035] ; In the monitoring environment of bird call sound patterns, background noise usually has a wide distribution frequency and a gentle energy fluctuation, showing a Gaussian-like characteristic, and its kurtosis is generally less than 3; while bird calls often present as short-time and sudden pulse-type signals, with obvious transient characteristics and high energy concentration, showing a steep rising edge and significant harmonic structure in the time-frequency domain, and its kurtosis is usually much greater than 3. By calculating the kurtosis of each IMF component, the bird call signal can be easily identified.

[0036] Furthermore, the principle of the MUSIC algorithm described in this invention is as follows: The theoretical basis of the two-dimensional MUSIC algorithm is the same as that of the one-dimensional MUSIC algorithm, but the two-dimensional algorithm adds the estimated parameter of the azimuth angle. The combination of multi-dimensional parameters can produce a gradually unbiased estimate, which is more suitable for practical applications.

[0037] Suppose the following conditions are met: (1) All signal sources have the same center frequency; (2) The acquisition array is a uniform circular array, the number of sensors is M, all sensors are the same and have the same performance in all directions; (4) The wavelength of the highest frequency signal is λ, and the spacing d of the microphone array is less than half of λ; (5) The received sound source signals are parallel to each other and regarded as plane waves; (6) The array elements are independent of the test signal; (7) The number of test signal sources is less than the number of sensors, that is, D < M; (8) The characteristics of the received signal branches are the same. The serial number of the signal source signal received by the microphone array is k ( k = 1, 2..., D), and the wavefront signal is Assuming it's a narrowband signal. By using the simplified conditions for narrowband signals, It can be represented as: ; in yes The complex envelope, yes Angular frequency. Center frequency of all signals. They are the same, therefore: ; Where c is the speed of sound in air (340 m / s), f is the frequency of the signal, and λ is the wavelength. Let the delay time between the array microphone units be... Under narrowband signal conditions, we have: ; Therefore, the delayed wavefront signal is: ; Taking a uniform circular array of M elements as an example, the center of the circle is the reference point of the array. At time t, the array elements... m ( m The response signal of the k-th signal source (e.g., = 1, 2, ..., M) is: ; in It is the influence coefficient. Since each element has no directionality ( =1). It is the relative delay from signal source k to array m relative to the array center, expressed as: ; Where r is the radius of the array, and These are the elevation and azimuth angles of the signal source k, respectively. It is the negative angle between the signal source and the positive x-axis (ranging from 0° to 360°), while It is the angle between the signal source and the positive z-axis of the array (ranging from 0° to 90°). Figure 3 The three-dimensional spatial model of a uniform circular array of M elements is shown. Considering the upper-level noise and incident wave from all signal sources, the output signal of the m-th array element is: ; in This is the measurement noise. Convert the above equation into vector form, and let... Then we have: ; in: ; ; ; Then, ; in: ; ; At this point, the problem becomes sampling. and from Estimate the direction of arrival (DOA) of the signal source k. The covariance matrix R of X is: ; in yes The conjugate transpose of . The signal and noise are uncorrelated, and the noise represents white noise at zero. The following results can be obtained: ; in yes The conjugate transpose of . yes The conjugate transpose of . yes The conjugate transpose of the signal. The correlation matrix (P) of the spatial signal and the correlation matrix of the noise. ) can be represented as follows: ; ; in This is the noise power, and I is the M-order identity matrix. For a uniform circular array, when... , When i ≠ j, each column of matrix A is independent. Therefore, if P is a non-singular matrix, then: ; If matrix P is positive definite, then All of R's eigenvalues ​​are positive. Therefore, R has M eigenvalues ​​(λ1, λ2, … λ). M ( ), is a full-rank matrix. Arranging the M eigenvalues ​​in descending order yields the following equation: ; Since the eigenvalues ​​corresponding to the signal are greater than the eigenvalues ​​corresponding to the noise, therefore λ1~λ D Let λ be the characteristic value of the signal. D+1 ~λ MLet λ be a noise eigenvalue. Therefore, the eigenvectors of R can also be divided into eigenvectors corresponding to the signal and eigenvectors corresponding to the noise. i It is the first of matrix R i There are eigenvalues, v i If is the corresponding eigenvector, then: ; if If is the smallest eigenvalue of R, then: ; ; The following permutations can be obtained: ; because It is a D×D dimensional full-rank matrix. The inverse matrix of P exists, and the inverse matrix of P... It also exists. Multiplying both sides of the above equation by... The following equation is obtained: ; therefore, ; That is, the column vectors of the noise eigenvectors and the signal matrix are perpendicular to each other. If a vector is found that is orthogonal (or nearly orthogonal) to the noise subspace, its direction represents the direction of wave arrival. The noise matrix is ​​constructed using the noise eigenvectors as columns.

[0038] ; in These are the MD eigenvectors corresponding to the noise. Then, the two-dimensional spatial spectrum is defined: ; Ideally, the denominator should be zero. However, due to the presence of noise in the actual bird-locating environment, the denominator has a minimum value. Therefore, there is a peak value. By searching in two-dimensional space, we can make θ and By observing changes in the sound spectrum and identifying the peak value in the two-dimensional spatial spectrum, the location of the sound source can be determined. Based on the relevant scenario, the steps for bird sound signature localization along a road are as follows: When birds around the road emit calls, the following covariance matrix can be estimated based on data received from N microphone array sensors.

[0039] ; in It is the i-th data received by the sensor. yes The conjugate transpose of the covariance matrix is ​​obtained. By decomposing this covariance matrix, the eigenvalues ​​and their corresponding eigenvectors can be obtained. Based on the order of the eigenvalues, D eigenvalues ​​and eigenvectors corresponding to the signal are constructed as the signal subspace, equal to the number of eigenvalues. The remaining MD eigenvectors form the noise subspace. Thus, the noise matrix can be obtained. Finally, traversing θ and θ in the space... The spectral function is calculated based on the equation. The maximum value of the two-dimensional spectral function can be found, and the azimuth and elevation angles of the sound source can be determined to locate the bird's call.

[0040] The algorithm described in this invention operates as follows: Step 1: Input a noisy bird call signal.

[0041] The system receives raw signals collected by a microphone array, which include bird calls and ambient noise, and represents these signals as a time series. x ( t ),in t This is a time point; the signal serves as the input parameter for the Variational Mode Decomposition (VMD) algorithm. f (i.e., the one-dimensional time-domain signal to be decomposed), used for subsequent noise suppression processing.

[0042] Step 2: Adaptively determine the number of VMD decomposition layers K.

[0043] Analyze the input signal x ( t The optimal number of decomposition layers is automatically calculated based on the spectral characteristics (calculation of spectral entropy and kurtosis). K ;Should K The value serves as the parameter for the number of decomposition modes in the VMD algorithm, controlling the precision of signal decomposition and ensuring that bird call features are effectively separated.

[0044] Step 3: Initialize VMD parameters.

[0045] Setting VMD core parameters: Penalty factor α (Adaptively adjusted based on signal bandwidth), convergence threshold tol (default 1×10) -6 Maximum number of iterations N (Default 500 times), Initial center frequency ωk (The signal frequency is obtained through FFT); these parameters are all input configurations for the VMD algorithm, used to optimize the decomposition process.

[0046] Step 4: Perform VMD decomposition.

[0047] Using the input signal from step 1 x ( t Step 2 decomposition layer KThe initialization parameters from step 3 are used to decompose the signal into its components using a variational mode decomposition algorithm. K Modal functions u k (k=1,2,…,K) k =1,2,…, K ); u k The output parameters of the VMD algorithm represent the decomposed modal components, including bird calls and noise components.

[0048] Step 5: Mode selection and signal reconstruction.

[0049] Calculate each module u k The correlation or kurtosis is used to select effective modes containing bird calls, and the selected modes are summed to obtain the reconstructed high signal-to-noise ratio signal, denoted as . x ^( t The reconstructed signal x ^( t () is used as the input signal for the Multiple Signal Classification (MUSIC) algorithm for sound source localization.

[0050] Step 6: Construct the array receive signal matrix.

[0051] The reconstructed signal output from step 5 x ^( t The signal is converted into a microphone array received signal matrix X, where each column of the matrix corresponds to the time-domain sampling data of a microphone channel; this matrix X serves as the input parameter for the MUSIC algorithm and is used for spatial spectrum estimation.

[0052] Step 7: Calculate the covariance matrix.

[0053] Calculate the covariance matrix for the array received signal matrix X. ,in M The number of sampling points. H This represents the conjugate transpose; the covariance matrix R is the core input parameter of the MUSIC algorithm, used for eigenvalue decomposition.

[0054] Step 8: Eigenvalue decomposition and subspace partitioning.

[0055] Perform eigenvalue decomposition on the covariance matrix R to obtain the eigenvalues. λ i and the corresponding feature vector v i Divide the signal subspace (corresponding to larger eigenvalues) and the noise subspace E according to the magnitude of the eigenvalues. n (The eigenvectors corresponding to the small eigenvalues); Noise subspace E n It is a key parameter of the MUSIC algorithm, used to construct the spatial spectrum function.

[0056] Step 9: Construct the MUSIC spatial spectrum.

[0057] Using the noise subspace E n and array manifold matrix a( θ ), calculate the spatial spectral function ,in θ The scanning angle; the spatial spectrum function P ( θ This is used for subsequent peak search to characterize the energy distribution in the direction of the sound source.

[0058] Step 10: Peak search and sound source localization.

[0059] In spatial spectrum P ( θ Search for the global peak value in the search bar; the angle corresponding to the peak value is the location of the bird call source. (Note: The last part is a typo and can be left as is.) ;Should The final output of the algorithm is the location parameters of the bird call source (DOA estimation result), which completes the localization of the bird sound source.

[0060] The improvements of this invention are as follows: Improvement 1: Adaptively determine the number of decomposition layers K.

[0061] Traditional VMD requires a preset number of decomposition layers K, which is subjective and prone to under- or over-decomposition. This method adaptively determines the optimal K value by gradually increasing the K value and monitoring changes in the reconstructed signal energy ratio, thus matching the number of decomposition layers to the actual characteristics of the signal and avoiding parameter dependence problems.

[0062] Improvement 2: IMF component screening mechanism (EKI screening).

[0063] Traditional kurtosis thresholding methods fail under complex noise conditions (such as Gaussian noise), easily resulting in noise residue or signal loss. This method proposes an EKI (Energy-Kurtosis Index) screening mechanism: IMF components are sorted in descending order of kurtosis value, and effective components are dynamically selected by combining energy accumulation (≥90%) and signal-to-noise ratio constraints (>3dB). This adaptively adapts to strong noise environments, ensuring accurate extraction of bird call signals.

[0064] Improvement point 3: IVMD and MUSIC cascade noise reduction design.

[0065] Traditional MUSIC positioning performance degrades sharply or even fails under low signal-to-noise ratio (SNR) conditions. This embodiment uses an IVMD as a pre-filter, leveraging its strong denoising capability to provide a high SNR input signal for the MUSIC, significantly improving positioning accuracy in low SNR environments (actual measurement shows a reduction in positioning error of over 40%).

[0066] Figure 4 This is a schematic diagram of the sound source localization using a uniform circular microphone array in this invention. Several microphones arranged in a circle are used to synchronously collect multi-channel acoustic signature signals from bird sound sources. In the diagram, the ray centered on the location of the bird sound source represents the direction of sound wave incidence. The angle of the incident direction relative to the horizontal plane is the azimuth angle, and the angle relative to the horizontal plane is the elevation angle. By using the time difference and phase difference of the signals received by each microphone, the spatial spectral peak can be determined within the two-dimensional search domain formed by the azimuth angle and the elevation angle, and the spatial location of the bird sound source can be determined accordingly, thus achieving accurate localization of bird activity near power transmission lines.

[0067] The proposed acoustic array-based bird localization method for power transmission lines exhibits significant comprehensive advantages compared to existing technologies. In terms of technical performance, this method achieves superior noise resistance through deep collaboration between improved VMD and MUSIC algorithms. Even in complex noise environments such as strong winds and equipment vibrations, it can accurately extract weak bird acoustic signature features, achieving a much higher localization accuracy than traditional acoustic localization methods. Simultaneously, its processing flow is highly efficient, meeting real-time monitoring requirements. In terms of engineering implementation, the system adopts a modular design. The core algorithm is built based on an acoustic physical model, eliminating the need for large-scale training datasets. It can be directly integrated into edge computing devices such as power transmission tower monitoring terminals, resulting in convenient deployment and low maintenance costs. The sensor array uses omnidirectional microphones equipped with professional protective housings, ensuring long-term stable operation in harsh outdoor environments with wide temperature ranges and high humidity. In terms of economic and social benefits, the large-scale application of this technology can significantly reduce the cost of traditional manual inspections. Through accurate early warning and timely bird removal, it can effectively reduce transmission line tripping accidents caused by bird damage and improve the safety and reliability of power grid operation. At the same time, it provides key technical support for building an intelligent and proactive transmission line operation and maintenance defense system, and powerfully promotes the transformation and upgrading of power grid operation and maintenance mode towards digitalization and intelligence.

[0068] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0069] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for locating birds on power transmission lines based on sound source localization, characterized in that, include: A uniform circular microphone array is deployed on the transmission tower to collect multi-channel raw acoustic data around the line, forming a noisy bird call signal. The noisy bird call signal is subjected to spectral analysis to calculate spectral entropy and kurtosis. Based on the criterion that the ratio of reconstructed energy to original energy tends to be stable, the number of decomposition levels of variational mode decomposition is adaptively determined and the variational mode decomposition parameters are initialized. The noisy bird call signal is decomposed into multiple intrinsic mode functions using a variational mode decomposition algorithm. The kurtosis or correlation of each intrinsic mode function is calculated to filter out effective modes containing bird calls. The effective modes are then superimposed and reconstructed to obtain a high signal-to-noise ratio bird call signal. The high signal-to-noise ratio bird call signal is constructed into a microphone array received signal matrix. The covariance matrix of the microphone array received signal matrix is ​​calculated and eigenvalue decomposition is performed to obtain the signal subspace and noise subspace. Based on the noise subspace and array manifold vector, a two-dimensional multi-signal classification spatial spectrum is constructed. Spectral peaks are searched in the two-dimensional search domain of azimuth and elevation angles, and the azimuth and elevation angles of the bird sound source corresponding to the spectral peaks are output to locate the bird's position.

2. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, The uniform circular microphone array satisfies the half-wavelength sampling condition of the wavelength corresponding to the highest frequency; the spacing between adjacent microphones in the uniform circular microphone array is less than half of the wavelength corresponding to the highest frequency of the monitored bird calls, and the number of sensors in the uniform circular microphone array is greater than the number of bird sound sources to be located.

3. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, The spectrum analysis includes: Perform a Fast Fourier Transform on the noisy bird call signal to obtain the dominant frequency distribution; The spectral entropy and kurtosis are calculated based on the dominant frequency distribution to drive the adaptive determination of the number of decomposition layers.

4. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, Adaptively determine the number of decomposition levels for variational mode decomposition, including: Starting from a preset initial candidate decomposition level, the number of levels is incremented by 1 each time. Modal decomposition and reconstruction are performed separately, and the ratio of the reconstructed signal energy to the original signal energy corresponding to each candidate decomposition level is calculated. When the energy ratio first stabilizes as the number of candidate decomposition layers increases, the corresponding number of candidate decomposition layers is determined as the number of decomposition layers.

5. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, The variational mode decomposition parameters include a penalty factor, a convergence threshold, a maximum number of iterations, and an initial center frequency; wherein the convergence threshold is 0.000001, the maximum number of iterations is 500, the initial center frequency is determined by the dominant frequency of the noisy bird call signal obtained by the fast Fourier transform, and the penalty factor is adaptively adjusted according to the signal bandwidth.

6. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, Filter valid modalities containing bird calls, including: Calculate the kurtosis of each intrinsic mode function and arrange them in descending order of kurtosis, then accumulate the corresponding energy percentages in sequence. When the cumulative energy percentage reaches or exceeds 90% and the estimated signal-to-noise ratio is greater than 3 dB, the corresponding number of intrinsic mode functions are determined as the effective modes.

7. The method for locating birds on power transmission lines based on sound source localization according to claim 6, characterized in that, The estimated signal-to-noise ratio is obtained by real-time estimation of the background noise energy corresponding to the remaining intrinsic mode functions that were not selected as the effective mode, and is used to adaptively adjust the dynamic truncation position of the effective mode.

8. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, The azimuth search range of the two-dimensional search domain is 0 degrees to 360 degrees, and the pitch search range is 0 degrees to 90 degrees. The global spectral peak is obtained as the azimuth and pitch output of the bird sound source by traversing the two-dimensional search domain.

9. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, After outputting the azimuth and elevation angles of the bird sound source, the azimuth and elevation angles of the bird sound source, the corresponding positioning time, and the high signal-to-noise ratio bird call signal are uploaded to the edge computing terminal device on the power transmission tower to realize real-time detection, positioning, and early warning linkage of bird activities.

10. The method for locating birds on power transmission lines based on sound source localization according to claim 1, characterized in that, The uniform circular microphone array uses omnidirectional microphones, and each omnidirectional microphone is equipped with a protective shell to ensure the long-term stable acquisition capability of the omnidirectional microphones in a wide temperature range and high humidity field environment.

Images