Charging pile noise monitoring method, noise monitoring system and readable medium

Abstract

This application relates to a charging pile noise monitoring method, a noise monitoring system, and a readable medium. The charging pile noise monitoring method utilizes a 3D spherical array for asynchronous measurement, collecting noise data from several measurement locations. The noise data is then processed for noise reduction and filtering to obtain a noise signal representing the target analysis frequency characteristic of the tested charging pile. This signal is then used to obtain a vector composed of beamforming energy estimates from all discrete scanning grid points, enabling rapid visualization of the sound source and improving the spatial resolution of sound source localization. Furthermore, a Bayesian inference algorithm is used to describe the variables in the sound field from a statistical optimization perspective, solving for the sound source energy distribution at discrete scanning grid points, accurately locating the fault location in the tested charging pile, with high imaging resolution. This allows for rapid localization of low-frequency noise generated by the tested charging pile, thus aiding in the rapid location of faults in DC fast charging piles.

Description

Technical Field

[0001] This application relates to the field of noise measurement technology, and in particular to a method for monitoring the noise of a charging pile, a noise monitoring system, and a readable medium. Background Technology

[0002] The existing charging piles vary in technological sophistication, but in order to meet the fast-paced lifestyle of modern cities, people want electric vehicles to charge as quickly as possible. Therefore, DC fast charging piles have a broad development prospect, and the design problems of DC fast charging piles are gradually emerging.

[0003] Currently, many DC fast charging stations have deficiencies in their noise control, and their improper placement in public places causes disturbance to residents. Related experiments have measured the noise level near a DC fast charging station during operation to be approximately 70 decibels, while near a residential building, the noise level is around 55 decibels, sounding like a large vacuum cleaner used in a shopping mall. In most cases, people are inside or near their electric vehicles while charging, and the noise generated by the DC fast charging station can easily bother them. Therefore, detecting and locating noise levels is of significant importance for manufacturers when designing and installing DC fast charging stations. The noise from DC fast charging stations is mainly generated by the cooling fan system and the transformer.

[0004] Currently, common detection methods include continuous scanning and discrete point measurement, both of which have their own limitations. Continuous scanning places extremely high demands on the control of the detection instrument, while discrete point measurement results in insufficient data samples, compromising accuracy. Conventional methods involve using noise meters for reasonable equipment testing. This approach can only provide a rough detection of high-frequency noise, offering little effective monitoring of low-frequency noise and making it difficult to pinpoint the noise source. Since most of the noise generated by DC fast charging piles is low-frequency, there are virtually no effective noise detection methods available for DC fast charging piles. Summary of the Invention

[0005] Therefore, it is necessary to provide a charging pile noise monitoring method, a noise monitoring system, and a readable medium to address the problem that conventional detection methods can only perform coarse detection of high-frequency noise, are almost unable to effectively monitor low-frequency noise, and are difficult to effectively locate the noise.

[0006] This application provides a charging pile noise monitoring method, applied to a noise monitoring system. The noise monitoring system includes a 3D sphere array, which contains multiple microphones. The charging pile noise monitoring method includes:

[0007] With the charging pile under test in the powered-on state, noise data is acquired at several measurement locations around the circumference of the charging pile under test; the noise data includes sound pressure signals measured by multiple microphones.

[0008] The acquired detection noise data is denoised to obtain an initial noise signal;

[0009] The initial noise signal is filtered to obtain the noise signal at the target analysis frequency;

[0010] Select a scanning area in the sound source plane and divide the scanning area into several discrete scanning grid points evenly;

[0011] Based on the noise signal of the target analysis frequency, obtain the vector composed of the beamforming energy estimates of all scan grid points;

[0012] An energy propagation model is established based at least on the vector formed by the beamforming energy estimates of all scanned grid points and the energy propagation matrix;

[0013] The high-resolution sound source energy distribution of the tested charging pile in the energy propagation model is solved based on the Bayesian inference algorithm.

[0014] This application also provides a noise monitoring system, including:

[0015] A 3D sphere array is used to collect noise data of the charging pile under test at the measurement location;

[0016] An information processing device, signal-connected to the 3D sphere array, includes a memory and a processor, the memory being coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the charging pile noise monitoring method as described above.

[0017] This application also provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the charging pile noise monitoring method as described above.

[0018] This application relates to a charging pile noise monitoring method, a noise monitoring system, and a readable medium. The charging pile noise monitoring method utilizes a 3D spherical array for asynchronous measurement, collecting noise data from several measurement locations. The noise data is then denoised and filtered to obtain a noise signal representing the target analysis frequency characteristic of the tested charging pile. This signal is then used to obtain a vector composed of beamforming energy estimates from all scanned grid points, enabling rapid visualization of the sound source inversion and improving the spatial resolution of sound source localization. The method exhibits good robustness and intuitiveness. Furthermore, a Bayesian inference algorithm is used to describe the variables in the sound field from a statistical optimization perspective, solving for the sound source energy distribution at the scanned grid points and accurately locating the fault location in the tested charging pile. The imaging results have high resolution, strong noise resistance, and strong adaptability. Therefore, the Bayesian-based beamforming sound source localization method can quickly direct low-frequency noise generated by the tested charging pile, thereby helping to quickly locate faults in DC fast charging piles. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a charging pile noise monitoring method provided in an embodiment of this application.

[0020] Figure 2 This is a schematic diagram showing the spatial distribution of the measurement location relative to the charging pile being measured in a charging pile noise monitoring method provided in an embodiment of this application.

[0021] Figure 3 This is a schematic diagram of the division of scanning grid points in a charging pile noise monitoring method provided in an embodiment of this application.

[0022] Figure 4 The 3D sphere array is positioned at the test point directly in front of the charging pile under test to collect noise signals and average noise spatial distribution maps.

[0023] Figure 5 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 30° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0024] Figure 6 The 3D sphere array is positioned at the test point, rotated 60° clockwise directly in front of the charging pile under test, to collect noise signals and the average noise spatial distribution map.

[0025] Figure 7 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 90° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0026] Figure 8 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 120° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0027] Figure 9 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 150° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0028] Figure 10 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 180° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0029] Figure 11 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 210° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0030] Figure 12 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 240° clockwise to test the noise signal and the average noise spatial distribution map.

[0031] Figure 13 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 270° clockwise to the test point to collect noise signals and average noise spatial distribution maps.

[0032] Figure 14 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 300° clockwise to test the noise signal and the average noise spatial distribution map.

[0033] Figure 15 The 3D sphere array is positioned directly in front of the charging pile under test and rotated 330° clockwise to test the noise signal and the average noise spatial distribution map.

[0034] Figure 16 The spatial distribution of noise constituted by all measurement locations provided in one embodiment of this application.

[0035] Figure 17 This is a schematic diagram illustrating the noise level classification of the spatial distribution of noise formed by all measurement locations provided in an embodiment of this application.

[0036] Figure 18 This is a sound source energy distribution map based on beamforming in a charging pile noise monitoring method provided in one embodiment of this application.

[0037] Figure 19 The charging pile noise monitoring method provided in one embodiment of this application is based on a Bayesian inference algorithm combined with a beamforming sound source energy distribution map. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0039] This application provides a method for monitoring noise in charging piles, applied to a noise monitoring system. The noise monitoring system includes a 3D sphere array, which contains multiple microphones. The 3D sphere array is prior art, and its specific details are not elaborated here.

[0040] like Figure 1 As shown, in one embodiment of this application, the charging pile noise monitoring method includes the following steps S100 to S700.

[0041] S100, with the charging pile under test in the powered-on state, acquires detection noise data at several measurement positions around the circumference of the charging pile under test.

[0042] Specifically, the detected noise data includes sound pressure signals measured by multiple microphones in the 3D sphere array.

[0043] S200 performs noise reduction processing on the acquired detection noise data to obtain the initial noise signal.

[0044] Specifically, due to the interference of background noise (such as other DC fast charging piles, pedestrians, vehicle noise, and shops) and environmental interference (such as temperature and air density) in the installation environment of the tested charging pile, this application performs noise reduction processing on the acquired detection noise data to eliminate the influence of the background environment and improve the accuracy of subsequent noise location results.

[0045] S300 filters the initial noise signal to obtain the noise signal at the target analysis frequency.

[0046] Specifically, the target analysis frequency can be a frequency range that represents the noise characteristics of the charging pile under test, or a specific frequency value or multiple specific frequency values.

[0047] More specifically, since most of the noise generated by DC fast charging piles is low-frequency, the target analysis frequency can be set to low frequency, which can effectively detect the noise of the charging pile under test.

[0048] The specific filtering method is to use a Bessel circuit to eliminate high-frequency noise from the initial noise signal and obtain low-frequency noise.

[0049] S400 selects a scanning area on the sound source plane and evenly divides the scanning area into several discrete scanning grid points.

[0050] Specifically, the sound source plane is the detection surface facing the 3D sphere array. The scanning area is the area centered on the charging pile under test, including the outer surface of the charging pile facing the 3D sphere array on one side, and can be scanned and detected by the 3D sphere array at various measurement positions in the circumferential direction.

[0051] For example, the following is an embodiment, such as Figure 3 As shown, the black dots represent discrete scanning grid points, evenly distributed across the scanning area. By dividing the scanning grid, the noise output of the charging pile under test is considered as the combined output of multiple discrete sound sources.

[0052] S500: Based on the noise signal of the target analysis frequency, obtain the vector composed of the beamforming energy estimates of all scan grid points.

[0053] S600, at least based on the vector formed by the beamforming energy estimates of all scanned grid points and the energy propagation matrix, establish an energy propagation model.

[0054] S700 uses a Bayesian inference algorithm to solve the high-resolution sound source energy distribution of the tested charging pile in the energy propagation model.

[0055] In this embodiment, asynchronous measurements are performed using a 3D spherical array to collect noise data at several measurement locations. The noise data is then denoised and filtered to obtain a noise signal representing the target analysis frequency characteristic of the tested charging pile. This signal is then used to obtain a vector composed of beamforming energy estimates from all scanned grid points, enabling rapid visualization of the sound source inversion and improving the spatial resolution of sound source localization. This method is robust and intuitive. Furthermore, a Bayesian inference algorithm is used to describe the variables in the sound field from a statistical optimization perspective, solving for the sound source energy distribution at the scanned grid points and accurately locating the fault location in the tested charging pile. This method offers high imaging resolution, strong noise resistance, and strong adaptability. Therefore, the Bayesian-based beamforming sound source localization method can quickly direct low-frequency noise generated by the tested charging pile, thus helping to quickly locate faults in DC fast charging piles.

[0056] By adopting the above methods, faults in DC fast charging piles can be detected and dealt with in a timely manner in their early stages, greatly reducing the probability of various accidents caused by the faults. When DC fast charging piles produce abnormal noises or sounds in abnormal frequency bands during operation, the problem can be reported in a timely manner, thereby preventing accidents from occurring. Furthermore, the structural design of the charging pile can be adjusted accordingly.

[0057] In one embodiment of this application, S100 includes the following S110 to S120.

[0058] S110 sets multiple measurement positions with the center of the outline of the charging pile under test as the origin and a preset distance as the radius, according to a preset angle interval.

[0059] Specifically, the outline of the outer surface of the charging pile device under test is drawn in a plane, and the center of the outline is used as the origin of the movement trajectory of the 3D ball array.

[0060] S120, at a preset measurement height and along a preset moving direction, collects detection noise at least once at each measurement position.

[0061] Specifically, such as Figure 2 As shown, while keeping the distance from the origin constant, the device moves at a clockwise (or counterclockwise) rotation angle of 30° to collect detection noise data at all measurement locations around the charging pile under test. Figure 2 The medium-dark sphere array pattern indicates the current measurement position of the 3D sphere array, while the lighter sphere array pattern indicates other measurement positions.

[0062] Of course, as another measurement method of this application, while keeping the distance from the outer surface of the charging pile under test unchanged, the device can move around the charging pile under test by rotating at a clockwise angle of 30° to obtain the detection noise data of all measurement positions.

[0063] The preset measurement height is the height of the cooling fan system or transformer of the charging pile under test, or half the height of the charging pile under test.

[0064] In this example, the cooling fan system and transformer of the charging pile under test are the main noise sources of the charging pile, so they are used as the main monitoring points.

[0065] In one embodiment of this application, S200 includes the following S210 to S230.

[0066] S210 performs a short-time Fourier transform on the detected noise data to obtain the original spectrum and the original phase.

[0067] S220, the original spectrum is sent to a noise prediction model, the noise prediction model is run, and the background noise spectrum output by the noise prediction model is obtained.

[0068] Specifically, the noise prediction model converges during training when the charging pile under test is not in operation. The specific construction method is as follows: Pure signals and noise signals are sampled at certain frequencies to form multiple new pure and noise signals of equal length. The pure and noise signals are then randomly scaled and mixed to obtain various mixed audio signals, ensuring diversity. Fourier transforms are performed on the various mixed audio signals to obtain the final input spectrum, which is then used to train the noise prediction model.

[0069] S230 subtracts the background noise spectrum from the original spectrum and obtains the initial noise signal after noise reduction through inverse Fourier transform.

[0070] In this embodiment, the noise reduction processing of the acquired detection noise data is performed to eliminate the influence of the background environment and improve the accuracy of subsequent noise localization results.

[0071] In one embodiment of this application, in S500, the vector formed by the beamforming energy estimates of all scan grid points is:

[0072] y = [y1(f), y2(f), ..., y n (f),···,y N (f)] T Formula 1

[0073] Among them, y n The beamforming result at the target analysis frequency f at the nth scan grid point is given by N, where N represents the total number of scan grid points and T represents the vector transpose.

[0074] In formula 1, y n (f) is represented as:

[0075]

[0076] Among them, a n Let be the steering vector from all microphones in the 3D sphere array to the nth scan grid point. The steering vector represents the inverse process of sound source propagation and plays a focusing role in beamforming. * denotes the conjugate transpose, and n represents the index of the scan grid point. ||·||2 is the L2 norm. This is the cross-power spectrum matrix.

[0077] In formula 2, a n (f) is represented as:

[0078] a n (f)=(a 1,n ,a 2,n ,···,a m,n ,…,a M,n ) T Formula 3

[0079] Where M represents the total number of microphones in the 3D sphere array, m is the serial number of the microphone in the 3D sphere array, and T represents the vector transpose.

[0080] In formula 3, a m,n Represented as:

[0081]

[0082] Where j is the complex unit, r m,n Let m be the distance between the m-th microphone and the n-th scan grid point, c be the speed of sound in the air, and f be the target analysis frequency.

[0083] In Formula 2 Represented as:

[0084]

[0085] Where p(f)=(p1(f),p2(f),···,p m (f),···,p M (f)) T * indicates conjugate transpose. It expresses expectation.

[0086] Where, p m (f) represents the frequency component f obtained after noise reduction and filtering of the sound pressure signal measured at the m-th microphone in the 3D sphere array, and T represents the vector transpose.

[0087] In practical calculations, the cross-power spectrum matrix can be used... Set the main diagonal elements to 0 to eliminate the effects of noise.

[0088] In this embodiment, the beamforming method using a 3D spherical array for asynchronous measurement can quickly achieve visualized sound source inversion, exhibiting good robustness and intuitiveness. At each measurement location, a scanning grid point associated with the direction of maximum noise can be obtained for the charging pile under test. This allows for the acquisition of the maximum noise location at all measurement locations around the charging pile, achieving a complementary effect through asynchronous measurement and further improving longitudinal resolution.

[0089] The beamforming methods described above can quickly achieve visualized sound source inversion; however, as... Figure 18 As shown, the imaging results are rather blurry, especially at low frequencies.

[0090] In one embodiment of this application, S600 includes the following S610 to S620.

[0091] S610, construct the energy propagation matrix as shown in Equation 2:

[0092]

[0093] in, For elements in H.

[0094] S620, the energy propagation model is constructed as follows:

[0095]

[0096] Where y is the vector composed of the beamforming energy estimates of all scanned grid points, and x is the high-resolution sound source energy distribution, which is an unknown variable to be solved. ∈=(∈1,∈2,…,∈ n ) T The model error is represented by T, which denotes the vector transpose. Let n1 = 1, ..., n, n2 = 1, ..., N be the elements in the energy propagation matrix H. This is the guide vector from the n1th scan grid point to the 3D sphere array; is the guide vector from the n2th scan grid point to the 3D sphere array; * indicates conjugate transpose.

[0097] In one embodiment of this application, S700 includes the following S710 to S730.

[0098] S710, through variational prior q1(x|γ) x ),q2(γ x ) and q3(γ ∈ To approximate the joint distribution p(x,γ) x ,γ ∈ |y)∝p(θ,y).

[0099] Where q1(·), q2(·), q3(·) and p(·) are all probability density functions, and ∝ represents positive correlation.

[0100] S720, based on mean-field theory, the variational prior parameters x (i.e., high-resolution sound source energy distribution), γ x and γ ∈ The variables are integrated into a single variable θ that satisfies q(θ)=q1(x)q2(γx)q3(γ ∈ ). Where γ x Let γ be the covariance matrix of x. ∈ The covariance matrix of ∈ can be kept as a constant.

[0101] Minimize the KL (Kullback-Leibler) divergence criterion as shown in Equation 8:

[0102]

[0103] S730, based on prior knowledge, the variational prior of x is expressed as a multidimensional Gaussian distribution with mean μ and variance Σ. Based on the sparsity of the Student's t-distribution and the Gaussian mixture property, γ... x and γ ∈ The variational prior is expressed as a product of the inverse gamma distribution:

[0104]

[0105] in, This indicates a normal distribution.

[0106] According to the KL divergence minimization criterion, the iterative process of x is as follows:

[0107]

[0108] Where l represents the l-th iteration.

[0109] γ x and γ ∈ The iterative process is as follows:

[0110]

[0111] The iteration cutoff condition is ρ is a constant representing the acceptable error.

[0112] In this embodiment, the Bayesian method can describe the variables in the sound field from a statistical optimization perspective, solve for the high-resolution sound source energy distribution x, and then combine the Bayesian inference algorithm with beamforming for sound source localization. This allows for rapid direction finding of low-frequency noise generated by DC fast charging piles, such as... Figure 19 As shown, its imaging results have high resolution, strong noise resistance, and strong adaptability.

[0113] In one embodiment of this application, after S700, the charging pile noise monitoring method further includes the following S810 to S820.

[0114] S810, calculate the average noise value of the tested charging pile at the measurement location based on the detected noise data.

[0115] S820 constructs a noise radiation map of all measurement locations based on the azimuth angles of all measurement locations and the noise value corresponding to each measurement location.

[0116] For example, one embodiment is described below. Figures 3 to 15 The acquired noise signals and average noise spatial distribution at each measurement location are shown, and the average noise values ​​are mapped to the corresponding measurement locations in the highest temperature spatial distribution map.

[0117] Figure 16 The spatial distribution of noise across all measurement locations is shown, including the average noise value across all measurement locations.

[0118] Figure 17 The noise level classification of the spatial distribution of noise across all measurement locations is shown. During the experiment, only the charging head on the left side of the tested charging pile was operational. Figure 17It can be seen that the higher noise levels are found at the air outlets on both sides of the charging pile, and the noise from the left air outlet has exceeded the Class 4 noise standard, causing noise pollution.

[0119] In this embodiment, by constructing a noise radiation map, the noise surroundings near the tested charging pile are visually displayed, which has important reference value for manufacturers in designing and installing DC fast charging piles.

[0120] In one embodiment of this application, in S100, multiple sets of detection noise data are collected at each measurement location.

[0121] S810 includes the following S811 to S815.

[0122] S811, Select a measurement location.

[0123] S812, determine whether the difference between the maximum and minimum values ​​in the multiple sets of detection noise data at the measurement location is greater than 5 dB.

[0124] S813, if the difference between the maximum and minimum values ​​in the multiple sets of detected noise data at the measurement location is greater than 5 dB, then the average value of the multiple sets of detected noise data is calculated using the energy averaging method, and is used as the average noise value at the measurement location.

[0125] Specifically, the energy averaging method calculates the mean by collecting 10 sets of detection noise data at the same test time and time interval. The specific calculation method can be referred to the ISO near-field test method, which will not be elaborated here.

[0126] S814, if the difference between the maximum and minimum values ​​in the multiple sets of detected noise data at the measurement location is less than or equal to 5 dB, then the arithmetic mean of the multiple sets of detected noise data is calculated as the average noise value at the measurement location.

[0127] S815, return to the step of selecting a measurement location until all measurement locations have been selected.

[0128] This application also provides a noise monitoring system.

[0129] In one embodiment of this application, the noise monitoring system includes a 3D spherical array and an information processing device.

[0130] Specifically, the 3D sphere array is used to acquire noise data of the charging pile under test at the measurement location. The 3D sphere array includes multiple microphones for acquiring sound pressure signals. More specifically, the 3D sphere array employs a high-performance 64-bit sphere array, which maximizes the accuracy of noise measurement, thereby highly simulating the noise interference that people might experience in a real-world environment.

[0131] An information processing device is signal-connected to the 3D sphere array and includes a memory and a processor, with the memory coupled to the processor. The memory stores program data, and the processor executes the program data to implement the charging pile noise monitoring method as described above.

[0132] This application also provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the charging pile noise monitoring method as described above.

[0133] The technical features of the above embodiments can be combined arbitrarily, and the execution order of the method steps is not restricted. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

Claims

1. A charging pile noise monitoring method applied to a noise monitoring system, the noise monitoring system comprising a 3D spherical array, the 3D spherical array comprising a plurality of microphones, characterized in that, The charging pile noise monitoring method includes: With the charging pile under test in the powered-on state, noise detection data are acquired at several measurement positions around the circumference of the charging pile. Multiple measurement positions are set with the center of the charging pile's outline as the origin and a preset distance as the radius, at preset angle intervals. At a preset measurement height and along a preset direction of movement, noise detection is collected at least once at each measurement position. The noise detection data includes sound pressure signals measured by multiple microphones. In the process of acquiring noise detection data at several measurement positions around the charging pile under test while it is powered-on, multiple sets of noise detection data are collected at each measurement position. The step of calculating the noise value of the tested charging pile at the measurement location based on the detected noise data includes: Select a measurement location; Determine whether the difference between the maximum and minimum values ​​in multiple sets of detected noise data at the measurement location is greater than 5 dB; If the difference between the maximum and minimum values ​​in multiple sets of detected noise data at the measurement location is greater than 5 dB, then the average value is calculated using the energy averaging method for the multiple sets of detected noise data, and this average value is taken as the noise value at the measurement location. If the difference between the maximum and minimum values ​​in multiple sets of detected noise data at the measurement location is less than or equal to 5 dB, then the arithmetic mean of the multiple sets of detected noise data is calculated and used as the noise value at the measurement location. Return to the previous step of selecting a measurement location, and continue until all measurement locations have been selected; The acquired detection noise data is denoised to obtain an initial noise signal; the detection noise data is then subjected to a short-time Fourier transform to obtain the original spectrum and original phase; the original spectrum is sent to a noise prediction model, which is then run to obtain the background noise spectrum output by the noise prediction model; the noise prediction model is trained and converged while the charging pile under test is not in operation; the background noise spectrum is subtracted from the original spectrum, and the denoised initial noise signal is obtained by inverse Fourier transform. The initial noise signal is filtered to obtain the noise signal at the target analysis frequency; Select a scanning area in the sound source plane and divide the scanning area into several discrete scanning grid points evenly; Based on the noise signal of the target analysis frequency, obtain the vector composed of the beamforming energy estimates of all scan grid points; An energy propagation model is established based at least on the vector formed by the beamforming energy estimates of all scanned grid points and the energy propagation matrix; The high-resolution sound source energy distribution of the tested charging pile in the energy propagation model is solved based on the Bayesian inference algorithm. The noise value of the tested charging pile at the measurement location is calculated based on the detected noise data. Based on the azimuth of all measurement locations and the noise value corresponding to each measurement location, a noise radiation map consisting of all measurement locations is constructed.

2. The charging pile noise monitoring method according to claim 1, characterized in that, In the vector formed by the beamforming energy estimates of all scanned grid points obtained based on the noise signal of the target analysis frequency, the vector formed by the beamforming energy estimates of all scanned grid points is: Formula 1 wherein is the beamforming result at the i-th scan grid point at the target analysis frequency N denotes the total number of scan grid points, is the beamforming result at the i-th scan grid point at the target analysis frequency denotes vector transposition; In the formula 1 is represented by: Formula 2 wherein is the steering vector of the th microphone in the 3D spherical array to the th scan grid point, is the steering vector of the th microphone in the 3D spherical array to the th scan grid point, denotes the conjugate transpose, and n denotes the index of the scan grid point; is the norm, is the norm, is the cross-power spectral matrix; In the formula 2 is represented as: Official 3 in, This represents the total number of microphones in the 3D sphere array, where m is the serial number of the microphone in the 3D sphere array. Indicates vector transpose; In Formula 3 Represented as: Official 4 Where j is the complex unit, The distance between the m-th microphone and the n-th scan grid point. The speed of sound in air. Analyze the frequency for the target; In Formula 2 Represented as: Official 5 in, ; in, Represents the 3D sphere array. The sound pressure signals measured at each microphone are obtained after noise reduction and filtering. Frequency components, Let T denote the expectation, and let T denote the vector transpose.

3. The charging pile noise monitoring method according to claim 2, characterized in that, The step of establishing the energy propagation model based at least on the vector formed by the beamforming energy estimates of all scanned grid points and the energy propagation matrix includes: Construct the energy propagation matrix as shown in Equation 6: Official 6 in, For elements in H; The energy propagation model is constructed as follows: Official 7 in, The vector formed by the beamforming energy estimates of all scanned grid points. For model error, This represents the transpose of a vector. Energy propagation matrix The elements in , , For the first The guide vector from each scan grid point to the 3D sphere array; For the first The guide vector from each scan grid point to the 3D sphere array; This indicates the conjugate transpose.

4. The charging pile noise monitoring method according to claim 3, characterized in that, The high-resolution sound source energy distribution of the tested charging pile in the energy propagation model, obtained using the Bayesian inference algorithm, includes: Through variational priors and Approximating the joint distribution ; in, , , and Both are probability density functions. Indicates a positive correlation; According to mean-field theory, the variational prior parameters , and Integrate into variables satisfy ;in for The covariance matrix, The covariance matrix; The criterion for minimizing the KL divergence is shown in Equation 8: Official 8 Based on prior knowledge, The variational prior is represented as having a mean of The variance is The multidimensional Gaussian distribution, based on the sparsity and Gaussian mixture properties of the Student's t-distribution, will... and The variational prior is expressed as a product of the inverse gamma distribution: , Official 9 in, Represents a normal distribution; According to the KL divergence minimization criterion The iterative process is as follows: Official 10 in, Indicates the first The next iteration; and The iterative process is as follows: Official 11 The iteration cutoff condition is , It is a constant representing the acceptable error.

5. A noise monitoring system, characterized in that, include: A 3D sphere array is used to acquire noise data of the charging pile under test at the measurement location; An information processing device, signal-connected to the 3D sphere array, includes a memory and a processor, the memory being coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the charging pile noise monitoring method as described in any one of claims 1 to 4.

6. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by the processor, it implements the charging pile noise monitoring method as described in any one of claims 1 to 4.

Images