Active noise suppression system and method and medium

Abstract

The invention relates to the technical field of environmental protection, and discloses an active noise suppression system and method and a medium. Comprising a time delay estimation module used for carrying out spectral analysis on collected noise signals, identifying the frequency of a noise source, filtering the noise signals of the noise source with the corresponding frequency by using a filter with a set filter coefficient to obtain filtered noise signals, and sending the filtered noise signals to a time delay estimation module; estimating the time lag between the filtered noise signal and the vibration signal by using cross-spectrum analysis; and the noise suppression calculation module is used for performing time delay compensation by using the trained time delay compensation neural network model according to the estimated time delay between the filtered noise signal and the vibration signal to obtain a time delay compensated vibration signal, and calculating the time delay compensated vibration signal by using a dynamic feedback compensation algorithm to obtain a noise suppression result. And obtaining a noise suppression signal. According to the method, the time delay compensation efficiency is effectively improved, a better compensation effect is achieved, and the real-time performance and accuracy of noise suppression are improved.

Description

Technical Field

[0001] This invention relates to the field of environmental protection technology, specifically to active noise suppression systems, methods, and media. Background Technology

[0002] With the continuous expansion of urban construction, the number of substations and transformers in urban residential and commercial areas is gradually increasing, and noise problems are also gradually expanding. This not only affects the tranquility of the surrounding environment but also has a negative impact on the physical and mental health of residents. Long-term exposure to high-noise environments may lead to hearing damage, sleep disorders, and increased stress. Furthermore, noise is often accompanied by mechanical vibrations and stress within the equipment; long-term noise exposure may lead to premature transformer failure and a shortened lifespan.

[0003] Currently, existing noise suppression methods fall into two categories: passive suppression and active suppression. Passive suppression mostly employs passive sound insulation measures, such as soundproof enclosures, sound barriers, and vibration-damping bases. However, this method suffers from drawbacks such as poor heat dissipation, limited noise reduction effect, high construction and maintenance costs, complex installation processes, and long installation times. In contrast to passive suppression methods, active noise cancellation (ANC) technology neutralizes noise by generating reverse sound waves, enabling flexible and precise noise suppression, especially low-frequency noise. However, despite the significant advantages of ANC technology, its system design and implementation also face challenges related to complexity and high cost. Furthermore, in environments with complex noise sources and strong interference, it still faces problems such as slow convergence speed and lag in system response. Summary of the Invention

[0004] The purpose of this invention is to overcome the problems of limited noise reduction effect, slow convergence speed and system response lag in existing noise suppression methods for substations and transformers in environments with complex noise sources and strong interference. This invention provides an active noise suppression system, method and medium that can solve the problems of poor real-time performance and accuracy of existing noise suppression methods.

[0005] To achieve the above objectives, the present invention provides an active noise suppression system, comprising:

[0006] The time delay estimation module is used to collect noise and vibration signals from noise sources in substations, perform spectrum analysis on the collected noise signals to identify the frequency of the noise sources, and use a filter with set filtering coefficients to filter the noise signals of the noise sources at the corresponding frequencies to obtain the filtered noise signals. Cross-spectrum analysis is then used to estimate the time delay between the filtered noise signals and the vibration signals.

[0007] The noise suppression calculation module is used to perform time delay compensation based on the estimated time delay between the filtered noise signal and the vibration signal, using a trained time delay compensation neural network model to obtain the time delay compensated vibration signal. The dynamic feedback compensation algorithm is then used to calculate the noise suppression signal from the time delay compensated vibration signal.

[0008] Furthermore, the method for identifying the frequency of the noise source by performing spectral analysis on the collected noise signal is as follows:

[0009]

[0010] Where X(f) represents the spectrum of the noise signal; x(n) represents the noise signal corresponding to the nth sampling point in the time domain; f represents the frequency of the noise; N represents the length of the noise signal; and j represents the complex operator.

[0011] Furthermore, the method for setting the filter coefficients is as follows:

[0012] The minimum mean square error adaptive filtering method adjusts the filter coefficients step by step according to the error between the input signal and the desired output. Its update formula is as follows:

[0013] w(k+1)=w(k)+μe(k)x(k)

[0014] Where w(k) represents the filter coefficient at time k; w(k+1) represents the filter coefficient at time k+1; μ represents the filter step size parameter; e(k) represents the error signal between the input signal and the desired output at time k; and x(k) represents the input signal at time k.

[0015] Furthermore, the method for estimating the time delay between the filtered noise signal and the vibration signal using cross-spectral analysis is as follows:

[0016] The cross-spectral density value between the noise signal and the vibration signal of the noise source is calculated using the following method:

[0017]

[0018] Among them, S xy (f) represents the cross-spectral density values ​​of x(n) and y(n); y * (n) represents the conjugate complex number of y(n); y(n) represents the vibration signal corresponding to the nth sampling point in the time domain;

[0019] Based on the cross-spectral density values ​​between the noise signal and the vibration signal from the noise source, the time delay τ(f) between the filtered noise signal and the vibration signal is estimated using the following method:

[0020]

[0021] Among them, ∠S xy (f) represents the cross-spectral density value S xy (f) corresponds to the phase value.

[0022] Furthermore, correlation analysis is used to estimate the correlation time delay between the filtered noise signal and the vibration signal, and the correlation time delay is used to train the time delay compensation neural network model to obtain the trained time delay compensation neural network model.

[0023] The method for estimating the correlation time delay between the filtered noise signal and the vibration signal using correlation analysis is as follows:

[0024] The correlation value between the noise signal and the vibration signal of the noise source is calculated using the following method:

[0025]

[0026] Where Corr(x(t),y(t+τ)) represents the correlation value between the noise signal x(t) and the vibration signal y(t+τ) at time t; Cov(x(t),y(t+τ)) represents the covariance between x(t) and y(t+τ); δ(x(t)) represents the variance of the noise signal x(t) at time t, and δ(y(t+τ)) represents the variance of the vibration signal y(t+τ) at time t;

[0027] The maximum correlation value is determined as the correlation lag, and the calculation method is as follows:

[0028] τ=argmax(Corr(x(t),y(t+τ)))

[0029] Where τ represents the correlation time delay between the filtered noise signal and the vibration signal.

[0030] Furthermore, the trained time-delay compensation neural network model is generated based on the interaction of the LSTM model and the GRU model;

[0031] Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using a trained time delay compensation neural network model to obtain the time delay-compensated vibration signal. The method is as follows: the estimated time delay between the filtered noise signal and the vibration signal is input into the LSTM model of the trained time delay compensation neural network model for first time delay compensation to obtain a first compensated vibration signal; the estimated time delay between the filtered noise signal and the vibration signal is input into the GRU model of the trained time delay compensation neural network model for second time delay compensation to obtain a second compensated vibration signal; weight values ​​are assigned to the first compensated vibration signal and the second compensated vibration signal respectively; and the time delay-compensated vibration signal is obtained based on the first compensated vibration signal, the second compensated vibration signal, the weight value of the first compensated vibration signal, and the weight value of the second compensated vibration signal.

[0032] Furthermore, the dynamic feedback compensation algorithm is based on a multi-channel feedback system, with each channel processing noise in different frequency bands;

[0033] The method for calculating the noise-suppressed signal using a dynamic feedback compensation algorithm on the vibration signal after time-delay compensation is as follows:

[0034]

[0035] Among them, u k Indicates noise suppression signal; x i (t) represents the vibration signal after time delay compensation from the input of the i-th channel; h i This represents the filter coefficients after the filter settings for the i-th channel;

[0036] Simultaneously, based on the noise signal frequency and time delay characteristics, the feedback gain is dynamically adjusted to optimize the control parameters of the dynamic feedback compensation algorithm, as follows:

[0037]

[0038] Where μ(k) represents the feedback gain and α represents the adjustment factor.

[0039] A second aspect of the present invention provides a method for active noise suppression, comprising:

[0040] Noise and vibration signals from noise sources in a substation are collected. The collected noise signals are subjected to spectrum analysis to identify the frequency of the noise source. The noise signals of the corresponding frequency noise sources are filtered using a filter with set filtering coefficients to obtain the filtered noise signals. Cross-spectrum analysis is used to estimate the time delay between the filtered noise signals and the vibration signals.

[0041] Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using the trained time delay compensation neural network model to obtain the time delay compensated vibration signal. The noise suppression signal is then calculated using a dynamic feedback compensation algorithm.

[0042] Furthermore, the method for identifying the frequency of the noise source by performing spectral analysis on the collected noise signal is as follows:

[0043]

[0044] Where X(f) represents the spectrum of the noise signal; x(n) represents the noise signal corresponding to the nth sampling point in the time domain; f represents the frequency of the noise; N represents the length of the noise signal; and j represents the complex operator.

[0045] Furthermore, the method for setting the filter coefficients is as follows:

[0046] The minimum mean square error adaptive filtering method adjusts the filter coefficients step by step according to the error between the input signal and the desired output. Its update formula is as follows:

[0047] w(k+1)=w(k)+μe(k)x(k)

[0048] Where w(k) represents the filter coefficient at time k; w(k+1) represents the filter coefficient at time k+1; μ represents the filter step size parameter; e(k) represents the error signal between the input signal and the desired output at time k; and x(k) represents the input signal at time k.

[0049] Furthermore, the method for estimating the time delay between the filtered noise signal and the vibration signal using cross-spectral analysis is as follows:

[0050] The cross-spectral density value between the noise signal and the vibration signal of the noise source is calculated using the following method:

[0051]

[0052] Among them, S xy (f) represents the cross-spectral density values ​​of x(n) and y(n); y * (n) represents the conjugate complex number of y(n); y(n) represents the vibration signal corresponding to the nth sampling point in the time domain;

[0053] Based on the cross-spectral density values ​​between the noise signal and the vibration signal from the noise source, the time delay τ(f) between the filtered noise signal and the vibration signal is estimated using the following method:

[0054]

[0055] Among them, ∠Sxy (f) represents the cross-spectral density value S xy (f) corresponds to the phase value.

[0056] Furthermore, correlation analysis is used to estimate the correlation time delay between the filtered noise signal and the vibration signal, and the correlation time delay is used to train the time delay compensation neural network model to obtain the trained time delay compensation neural network model.

[0057] The method for estimating the correlation time delay between the filtered noise signal and the vibration signal using correlation analysis is as follows:

[0058] The correlation value between the noise signal and the vibration signal of the noise source is calculated using the following method:

[0059]

[0060] Where Corr(x(t),y(t+τ)) represents the correlation value between the noise signal x(t) and the vibration signal y(t+τ) at time t; Cov(x(t),y(t+τ)) represents the covariance between x(t) and y(t+τ); δ(x(t)) represents the variance of the noise signal x(t) at time t, and δ(y(t+τ)) represents the variance of the vibration signal y(t+τ) at time t;

[0061] The maximum correlation value is determined as the correlation lag, and the calculation method is as follows:

[0062] τ=argmax(Corr(x(t),y(t+τ)))

[0063] Where τ represents the correlation time delay between the filtered noise signal and the vibration signal.

[0064] Furthermore, the trained time-delay compensation neural network model is generated based on the interaction of the LSTM model and the GRU model;

[0065] Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using a trained time delay compensation neural network model to obtain the time delay-compensated vibration signal. The method is as follows: the estimated time delay between the filtered noise signal and the vibration signal is input into the LSTM model of the trained time delay compensation neural network model for first time delay compensation to obtain a first compensated vibration signal; the estimated time delay between the filtered noise signal and the vibration signal is input into the GRU model of the trained time delay compensation neural network model for second time delay compensation to obtain a second compensated vibration signal; weight values ​​are assigned to the first compensated vibration signal and the second compensated vibration signal respectively; and the time delay-compensated vibration signal is obtained based on the first compensated vibration signal, the second compensated vibration signal, the weight value of the first compensated vibration signal, and the weight value of the second compensated vibration signal.

[0066] Furthermore, the dynamic feedback compensation algorithm is based on a multi-channel feedback system, with each channel processing noise in different frequency bands;

[0067] The method for calculating the noise-suppressed signal using a dynamic feedback compensation algorithm on the vibration signal after time-delay compensation is as follows:

[0068]

[0069] Among them, u k Indicates noise suppression signal; x i (t) represents the vibration signal after time delay compensation from the input of the i-th channel; h i This represents the filter coefficients after the filter settings for the i-th channel;

[0070] Simultaneously, based on the noise signal frequency and time delay characteristics, the feedback gain is dynamically adjusted to optimize the control parameters of the dynamic feedback compensation algorithm, as follows:

[0071]

[0072] Where μ(k) represents the feedback gain and α represents the adjustment factor.

[0073] A third aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described above.

[0074] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0075] In this invention, when suppressing transformer noise, firstly, for any noise source of the transformer, the time delay between the noise of any noise source and the feedback signal output by the multi-frequency dynamic feedback controller can be determined; wherein, the noise source includes low-frequency noise sources and high-frequency noise sources; then, the time delay can be input into a trained network model for time delay compensation and a driving signal can be generated; wherein, the trained network model is generated based on an LSTM model and a GRU model; finally, the vibration signal generated by the inertial actuator can be driven according to the driving signal to actively suppress the noise of the transformer; wherein, the vibration signal is opposite to the vibration direction of the transformer.

[0076] Based on this, in this invention, by employing a pre-trained network model for time delay compensation, compared to existing technologies, this invention not only effectively improves the efficiency of time delay compensation but also achieves better compensation results, thereby significantly improving the real-time performance and accuracy of noise suppression. Furthermore, since this invention can actively suppress both low-frequency and high-frequency noise sources—that is, it can simultaneously suppress noise in different frequency bands in real time—compared to existing technologies, it can more accurately adapt to the characteristics of multi-frequency noise when processing it. This greatly overcomes the shortcomings of existing technologies in handling time delays and noise frequency bands, achieving a more efficient and stable noise suppression effect. It has strong real-time optimization capabilities and low implementation costs, and can significantly improve the overall performance of substation noise control. Attached Figure Description

[0077] Figure 1 This is a schematic diagram of the active noise suppression system of the present invention;

[0078] Figure 2 This is a schematic diagram of the framework of the time delay compensation neural network model trained by this invention;

[0079] Figure 3 This is a schematic diagram of the inertial actuator of the present invention;

[0080] Figure 4 This is a schematic diagram of the active noise suppression device of the present invention;

[0081] Figure 5 This is a schematic diagram of the active noise suppression hardware system of the present invention;

[0082] Figure 6 This is a flowchart illustrating the active noise suppression method of the present invention.

[0083] Figure Labels

[0084] 10-Active noise suppression device, 101-Processor, 102-Memory, 103-I / O interface, 104-Database. Detailed Implementation

[0085] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0086] The endpoints and any values ​​of the ranges disclosed herein are not limited to the precise ranges or values, and these ranges or values ​​should be understood to include values ​​close to these ranges or values. For numerical ranges, the endpoint values ​​of the various ranges, the endpoint values ​​of the various ranges and individual point values, and individual point values ​​can be combined with each other to obtain one or more new numerical ranges, which should be considered as specifically disclosed herein.

[0087] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions provided in the various embodiments of this invention can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0088] Example 1

[0089] like Figure 1 The active noise suppression system shown includes:

[0090] The time delay estimation module is used to collect noise and vibration signals from noise sources in substations, perform spectrum analysis on the collected noise signals to identify the frequency of the noise sources, and use a filter with set filtering coefficients to filter the noise signals of the noise sources at the corresponding frequencies to obtain the filtered noise signals. Cross-spectrum analysis is then used to estimate the time delay between the filtered noise signals and the vibration signals.

[0091] The noise suppression calculation module is used to perform time delay compensation based on the estimated time delay between the filtered noise signal and the vibration signal, using a trained time delay compensation neural network model to obtain the time delay compensated vibration signal. The dynamic feedback compensation algorithm is then used to calculate the noise suppression signal from the time delay compensated vibration signal.

[0092] When actively suppressing noise in a transformer, the present invention firstly deploys signal acquisition devices (e.g., noise sensors, vibration sensors) within the substation; then, the signal acquisition devices can be used to collect noise and vibration signals from different frequency ranges of the transformer, focusing primarily on the vibration characteristics of the transformer, cooling equipment, and other equipment, in order to obtain multi-frequency transformer noise.

[0093] Furthermore, since the layout of the signal acquisition devices and the design of the feedback loop directly affect the noise suppression effect and the system response speed, in this invention, to effectively suppress noise within the substation, the location of the signal acquisition devices is rationally arranged according to the distribution characteristics of the noise sources. Specifically, the signal acquisition devices should be placed around the transformer and its cooling device, as well as on the walls surrounding the transformer, to accurately monitor the frequency and time delay characteristics of the noise sources. Simultaneously, the signal acquisition devices should avoid mutual interference and be able to capture noise data from different locations, forming a spatially uniform coverage.

[0094] Furthermore, to ensure comprehensive noise characteristics are captured, multiple signal acquisition devices should be deployed within the substation, forming multiple monitoring points. Data from these acquisition devices will be used to obtain the noise distribution characteristics at different locations and frequency bands. The sampling frequency can be set to 5000Hz to obtain a clear spectrum.

[0095] In order to accurately locate each noise source, in this embodiment of the invention, spectrum analysis technology can be used to process the collected multi-frequency transformer noise in order to identify the main frequency band of the noise.

[0096] Specifically, Fast Fourier Transform (FFT) can be used to perform spectral analysis on multi-frequency transformer noise to convert the time-domain signal into a frequency-domain signal, thereby identifying low-frequency and high-frequency noise sources (transformer noise is basically high-frequency noise, so the signal needs to be processed and analyzed to extract useful signals to suppress vibrations at specific frequencies).

[0097] Therefore, the specific method for identifying the frequency of the noise source by performing spectrum analysis on the collected noise signal in this invention is as follows:

[0098]

[0099] Where X(f) represents the spectrum of the noise signal; x(n) represents the noise signal corresponding to the nth sampling point in the time domain; f represents the frequency of the noise; N represents the length of the noise signal; and j represents the complex operator.

[0100] In this invention, the noise sources specifically include low-frequency noise sources and high-frequency noise sources. Furthermore, to ensure effective noise reduction across all frequency bands of the noise spectrum, an adaptive filter can be used for adaptive filtering of the multi-frequency transformer noise.

[0101] First, the Least-Mean-Square (LMS) algorithm can be used to adaptively filter noise from any noise source to obtain the filtered noise. In practical applications, the filter coefficients need to be dynamically adjusted (the LMS algorithm can gradually adjust the filter coefficients based on the error between the input signal and the desired output, requiring adaptive noise control to change the filter bandwidth, etc.) to cope with the multi-frequency characteristics of noise sources in substations, thereby ensuring effective noise reduction across all frequency bands of the noise spectrum. Corresponding filters are set according to the different frequencies of noise sources (usually determined empirically), and the filters with set coefficients are used to filter the noise signals of the corresponding frequency noise sources to remove the main noise bands.

[0102] Therefore, the specific method for setting the filter coefficients is as follows:

[0103] The minimum mean square error adaptive filtering method adjusts the filtering coefficients progressively based on the error between the input signal (noise signal before filtering) and the desired output (noise signal after filtering; typically, the desired output is 0 to ensure noise suppression). The update formula is as follows:

[0104] w(k+1)=w(k)+μe(k)x(k)

[0105] Where w(k) represents the filter coefficient at time k; w(k+1) represents the filter coefficient at time k+1, i.e., the adjusted filter coefficient (the set filter coefficient); μ represents the filter step size parameter; e(k) represents the error signal between the input signal and the desired output at time k; and x(k) represents the input signal at time k.

[0106] The filter used in this invention is a Butterworth filter.

[0107] Next, the time delay characteristics between the noise source and the vibration signal can be analyzed in order to compensate for the time delay. That is, the time delay between the filtered noise signal and the vibration signal can be determined in order to compensate for the time delay.

[0108] This invention employs cross-spectral analysis to estimate time delay. Specifically, the method for estimating the time delay between the filtered noise signal and the vibration signal using cross-spectral analysis is as follows:

[0109] First, the cross-spectral density value between the noise signal and the vibration signal of the noise source can be calculated according to the preset cross-spectral density formula. The calculation method is as follows:

[0110]

[0111] Among them, S xy (f) represents the cross-spectral density values ​​of x(n) and y(n); y * (n) represents the conjugate complex number of y(n); y(n) represents the vibration signal corresponding to the nth sampling point in the time domain.

[0112] Based on the cross-spectral density values ​​between the noise signal and the vibration signal from the noise source, the time delay τ(f) between the filtered noise signal and the vibration signal is estimated using the following method:

[0113]

[0114] Among them, ∠S xy (f) represents the cross-spectral density value S xy (f) corresponds to the phase value.

[0115] The training method for the trained time delay compensation neural network model is as follows: the correlation delay between the filtered noise signal and the vibration signal is estimated using correlation analysis, and the time delay compensation neural network model is trained using the correlation delay to obtain the trained time delay compensation neural network model.

[0116] Furthermore, the method for estimating the correlation time delay between the filtered noise signal and the vibration signal using correlation analysis is as follows:

[0117] First, the correlation value between the noise signal and the vibration signal of the noise source can be calculated according to a preset cross-correlation function. The calculation method is as follows:

[0118]

[0119] Where Corr(x(t),y(t+τ)) represents the correlation value between the noise signal x(t) and the vibration signal y(t+τ) at time t; Cov(x(t),y(t+τ)) represents the covariance between x(t) and y(t+τ); δ(x(t)) represents the variance of the noise signal x(t) at time t, and δ(y(t+τ)) represents the variance of the vibration signal y(t+τ) at time t;

[0120] The maximum correlation value is determined as the correlation lag, and the calculation method is as follows:

[0121] τ=argmax(Corr(x(t),y(t+τ)))

[0122] Where τ represents the correlation time delay between the filtered noise signal and the vibration signal.

[0123] In this invention, the process of training the time delay compensation neural network model using correlation time delay includes: taking the existing filtered noise signal and vibration signal as input, estimating the correlation time delay τ between the filtered noise signal and vibration signal as the target output, and then training.

[0124] In this invention, the trained time-delay compensation neural network model is generated based on an LSTM model and a GRU model, such as... Figure 2 As shown.

[0125] In this invention, the method for obtaining a time-delay-compensated vibration signal by performing time-delay compensation using a trained time-delay compensation neural network model based on the estimated time delay between the filtered noise signal and the vibration signal is as follows: the estimated time delay τ(f) between the filtered noise signal and the vibration signal is input into the LSTM model of the trained time-delay compensation neural network model for first time-delay compensation to obtain a first compensated vibration signal; the estimated time delay τ(f) between the filtered noise signal and the vibration signal is input into the GRU model of the trained time-delay compensation neural network model for second time-delay compensation to obtain a second compensated vibration signal; the first compensated vibration signal and the second compensated vibration signal are assigned the same weight value respectively; and the time-delay-compensated vibration signal is obtained based on the first compensated vibration signal, the second compensated vibration signal, the weight value of the first compensated vibration signal, and the weight value of the second compensated vibration signal.

[0126] In a preferred embodiment, a noise suppression signal can be obtained by using a multi-channel feedback system to calculate multi-frequency, multi-channel feedback control signals based on the vibration signal after time delay compensation.

[0127] Furthermore, the dynamic feedback compensation algorithm is based on a multi-channel feedback system, with each channel processing noise in different frequency bands. The method for calculating the noise-suppressed signal using the dynamic feedback compensation algorithm on the time-delay-compensated vibration signal is as follows:

[0128]

[0129] Among them, u k Indicates noise suppression signal; x i (t) represents the vibration signal after time delay compensation from the input of the i-th channel; h i This represents the filter coefficients after the filter settings for the i-th channel.

[0130] Furthermore, this invention dynamically adjusts the feedback gain based on the noise signal frequency and time delay characteristics, optimizing the control parameters of the dynamic feedback compensation algorithm to ensure more stable calculations. The adjustment strategy is as follows:

[0131]

[0132] Where μ(k) represents the feedback gain; α represents the adjustment factor (generally taken between 0 and 1 based on experience).

[0133] Furthermore, the vibration of the noise source can be adjusted in real time through a dynamic feedback loop. Based on the time delay relationship between the vibration signal and the noise source, the noise suppression signal can be adjusted in a timely manner, thereby ensuring the effective suppression of multi-frequency noise.

[0134] After obtaining the noise suppression signal, the driver can be controlled by the noise suppression signal to generate a drive signal. The vibration signal generated by the inertial actuator is driven by the drive signal to actively suppress the noise of the transformer. In a specific implementation, the vibration signal is opposite to the vibration direction of the transformer.

[0135] Specifically, the inertial actuator of the present invention is as follows: Figure 3 As shown, the inertial actuator consists of a mass block, a spring, a frame (shell), and a motor. The mass block is connected to the spring, the motor's mover is connected to the mass block, and the motor's stator is connected to the frame. The mass block, made of metal, is used to generate inertial force. The spring provides elastic support and is used to adjust the system's natural frequency. The motor generates active actuation force to drive the mass block to move, thereby counteracting the vibration of the transformer shell.

[0136] Therefore, when actively suppressing noise in a transformer by driving an inertial actuator (e.g., a piezoelectric actuator or a magnetostrictive actuator) to generate vibration signals based on a drive signal, the motor in the inertial actuator can be driven to generate an actuating force based on the drive signal; then, based on the actuating force, the mass block in the inertial actuator is driven by a spring to generate a vibration signal; finally, the transformer can be actively suppressed based on the vibration signal.

[0137] In some specific embodiments of the present invention, the above-mentioned active noise suppression operation can be implemented in application scenarios such as the active noise suppression device 10. The active noise suppression device 10 can be used for active noise suppression in substations, transformers, etc., and can be, for example, a personal computer (PC), server, or laptop. Figure 4As shown, the noise active suppression device 10 may include one or more processors 101, memory 102, I / O interface 103, and database 104. Specifically, the processor 101 may be a central processing unit (CPU) or a digital processing unit, etc. The memory 102 may be volatile memory, such as random-access memory (RAM); the memory 102 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or the memory 102 may be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory 102 may be a combination of the above-mentioned memories. The memory 102 may store some program instructions for the noise active suppression operation provided by the present invention. When these program instructions are executed by the processor 101, they can be used to implement the steps of the noise active suppression operation provided by the present invention to effectively suppress noise from substations and transformers. Database 104 can be used to store data such as multi-frequency transformer noise, noise from various noise sources, vibration signals, trained time-delay compensation neural network models, driving signals, vibration signals, frequency domain block minimum mean square FBLMS algorithm, and cross-spectral density values ​​involved in the solution provided by the present invention.

[0138] In this invention, the active noise suppression device 10 can acquire noise acquisition commands through the I / O interface 103. Then, the processor 101 of the active noise suppression device 10 will effectively suppress the noise of the substation and transformer according to the noise active suppression operation program instructions provided by this invention in the memory 102. In addition, the noise of the multi-frequency transformer, the noise of each noise source, vibration signals, the trained time delay compensation neural network model, drive signals, vibration signals, frequency domain block least mean square FBLMS algorithm, and cross-spectral density values ​​can be stored in the database 104.

[0139] The technical solution of the present invention is not limited to Figure 4 The application scenarios shown can also be used in other possible application scenarios.

[0140] In a specific implementation, the hardware system for the active noise suppression operation described above in this invention can be as follows: Figure 5 As shown, the system includes a signal acquisition unit, an adaptive filter, a time delay compensation module, a multi-frequency dynamic feedback controller, multiple drivers, multiple inertial actuators, and a transformer.

[0141] Furthermore, when performing noise suppression based on this active noise suppression operation, firstly, a signal acquisition device can be used to acquire noise from the transformer to obtain a multi-frequency transformer noise signal. Then, an adaptive filter can be used to adaptively filter the multi-frequency transformer noise signal to obtain a filtered multi-frequency transformer noise signal. Next, based on the filtered multi-frequency transformer noise signal, a multi-frequency dynamic feedback controller can be used to estimate the time delay and determine the time delay between the signal source and the vibration signal. Then, based on this time delay, a time delay compensation module can be used to compensate for the time delay to obtain a compensated vibration signal. Then, the compensated vibration signal can be input into the multi-frequency dynamic feedback controller to obtain a multi-dimensional noise suppression signal. Next, based on this multi-dimensional noise suppression, multiple drivers are controlled to generate corresponding drive signals. Then, according to these drive signals, multiple inertial actuators can generate vibration signals to actively suppress noise from the transformer.

[0142] In one possible implementation, a real-time embedded controller (such as a hardware platform based on DSP, FPGA, or ARM architecture) or a programmable logic controller (PLC) system can be used to process noise data, calculate control signals, and drive actuators. These controllers support high-speed data processing and can receive signals from noise and vibration sensors in real time and convert them into control signals.

[0143] In some possible implementations, various aspects of the invention may also be implemented as a program component comprising program code that, when run on a computer device, causes the computer device to perform the steps of the methods described above according to various exemplary embodiments of the present application, such as the computer device performing the operations described above.

[0144] The implementation of all or part of the above steps can be accomplished by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps included in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. Alternatively, if the integrated unit of the present invention is implemented as a software functional module and sold or used as an independent part, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software part. This computer software part is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0145] In summary, the present invention has the following advantages:

[0146] By introducing a well-trained time-delay compensation neural network model and adaptive filtering for time-delay compensation, the present invention accurately estimates and compensates for the time delay between the noise source and the vibration signal. Therefore, compared with the prior art, the present invention not only solves the problems of delay and inaccurate noise suppression caused by time delay, but also significantly improves the system response speed and stability, and optimizes the real-time performance and accuracy of noise suppression.

[0147] Because this invention can suppress noise in different frequency bands in real time, compared with the prior art, it can more accurately adapt to the characteristics of the noise when dealing with multi-frequency noise. It overcomes the shortcomings of the prior art in terms of time delay and noise frequency band processing, and achieves a more efficient and stable noise suppression effect. It has strong real-time optimization capability and low implementation cost, and can significantly improve the overall performance of substation noise control.

[0148] Because this invention can continuously adjust control parameters based on real-time noise signals to ensure the system always maintains optimal operating conditions, it avoids manual intervention. Therefore, compared to existing passive sound insulation methods (such as soundproof enclosures and sound barriers), this invention not only significantly reduces implementation costs by not relying on large physical structures and complex sound insulation projects, but also can be directly applied to existing facilities through the integration of a real-time control system, reducing infrastructure modifications and lowering initial investment and maintenance costs.

[0149] Example 2

[0150] like Figure 6 The active noise suppression method shown includes:

[0151] Noise and vibration signals from noise sources in a substation are collected. The collected noise signals are subjected to spectrum analysis to identify the frequency of the noise source. The noise signals of the corresponding frequency noise sources are filtered using a filter with set filtering coefficients to obtain the filtered noise signals. Cross-spectrum analysis is used to estimate the time delay between the filtered noise signals and the vibration signals.

[0152] Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using the trained time delay compensation neural network model to obtain the time delay compensated vibration signal. The noise suppression signal is then calculated using a dynamic feedback compensation algorithm.

[0153] When actively suppressing noise in a transformer, the present invention firstly deploys signal acquisition devices (e.g., noise sensors, vibration sensors) within the substation; then, the signal acquisition devices can be used to collect noise and vibration signals from different frequency ranges of the transformer, focusing primarily on the vibration characteristics of the transformer, cooling equipment, and other equipment, in order to obtain multi-frequency transformer noise.

[0154] Furthermore, since the layout of the signal acquisition devices and the design of the feedback loop directly affect the noise suppression effect and the system response speed, in this invention, to effectively suppress noise within the substation, the location of the signal acquisition devices is rationally arranged according to the distribution characteristics of the noise sources. Specifically, the signal acquisition devices should be placed around the transformer and its cooling device, as well as on the walls surrounding the transformer, to accurately monitor the frequency and time delay characteristics of the noise sources. Simultaneously, the signal acquisition devices should avoid mutual interference and be able to capture noise data from different locations, forming a spatially uniform coverage.

[0155] Furthermore, to ensure comprehensive noise characteristics are captured, multiple signal acquisition devices should be deployed within the substation, forming multiple monitoring points. Data from these acquisition devices will be used to obtain the noise distribution characteristics at different locations and frequency bands. The sampling frequency can be set to 5000Hz to obtain a clear spectrum.

[0156] In order to accurately locate each noise source, in this embodiment of the invention, spectrum analysis technology can be used to process the collected multi-frequency transformer noise in order to identify the main frequency band of the noise.

[0157] Specifically, Fast Fourier Transform (FFT) can be used to perform spectral analysis on multi-frequency transformer noise to convert the time-domain signal into a frequency-domain signal, thereby identifying low-frequency and high-frequency noise sources (transformer noise is basically high-frequency noise, so the signal needs to be processed and analyzed to extract useful signals to suppress vibrations at specific frequencies).

[0158] Therefore, the specific method for identifying the frequency of the noise source by performing spectrum analysis on the collected noise signal in this invention is as follows:

[0159]

[0160] Where X(f) represents the spectrum of the noise signal; x(n) represents the noise signal corresponding to the nth sampling point in the time domain; f represents the frequency of the noise; N represents the length of the noise signal; and j represents the complex operator.

[0161] In this invention, the noise sources specifically include low-frequency noise sources and high-frequency noise sources. Furthermore, to ensure effective noise reduction across all frequency bands of the noise spectrum, an adaptive filter can be used for adaptive filtering of the multi-frequency transformer noise.

[0162] First, the Least-Mean-Square (LMS) algorithm can be used to adaptively filter noise from any noise source to obtain the filtered noise. In practical applications, the filter coefficients need to be dynamically adjusted (the LMS algorithm can gradually adjust the filter coefficients based on the error between the input signal and the desired output, requiring adaptive noise control to change the filter bandwidth, etc.) to cope with the multi-frequency characteristics of noise sources in substations, thereby ensuring effective noise reduction across all frequency bands of the noise spectrum. Corresponding filters are set according to the different frequencies of noise sources (usually determined empirically), and the filters with set coefficients are used to filter the noise signals of the corresponding frequency noise sources to remove the main noise bands.

[0163] Therefore, the specific method for setting the filter coefficients is as follows:

[0164] The minimum mean square error adaptive filtering method adjusts the filtering coefficients progressively based on the error between the input signal (noise signal before filtering) and the desired output (noise signal after filtering; typically, the desired output is 0 to ensure noise suppression). The update formula is as follows:

[0165] w(k+1)=w(k)+μe(k)x(k)

[0166] Where w(k) represents the filter coefficient at time k; w(k+1) represents the filter coefficient at time k+1, i.e., the adjusted filter coefficient (the set filter coefficient); μ represents the filter step size parameter; e(k) represents the error signal between the input signal and the desired output at time k; and x(k) represents the input signal at time k.

[0167] The filter used in this invention is a Butterworth filter.

[0168] Next, the time delay characteristics between the noise source and the vibration signal can be analyzed in order to compensate for the time delay. That is, the time delay between the filtered noise signal and the vibration signal can be determined in order to compensate for the time delay.

[0169] This invention employs cross-spectral analysis to estimate time delay. Specifically, the method for estimating the time delay between the filtered noise signal and the vibration signal using cross-spectral analysis is as follows:

[0170] First, the cross-spectral density value between the noise signal and the vibration signal of the noise source can be calculated according to the preset cross-spectral density formula. The calculation method is as follows:

[0171]

[0172] Among them, S xy (f) represents the cross-spectral density values ​​of x(n) and y(n); y * (n) represents the conjugate complex number of y(n); y(n) represents the vibration signal corresponding to the nth sampling point in the time domain.

[0173] Based on the cross-spectral density values ​​between the noise signal and the vibration signal from the noise source, the time delay τ(f) between the filtered noise signal and the vibration signal is estimated using the following method:

[0174]

[0175] Among them, ∠S xy (f) represents the cross-spectral density value S xy (f) corresponds to the phase value.

[0176] The training method for the trained time delay compensation neural network model is as follows: the correlation delay between the filtered noise signal and the vibration signal is estimated using correlation analysis, and the time delay compensation neural network model is trained using the correlation delay to obtain the trained time delay compensation neural network model.

[0177] Furthermore, the method for estimating the correlation time delay between the filtered noise signal and the vibration signal using correlation analysis is as follows:

[0178] First, the correlation value between the noise signal and the vibration signal of the noise source can be calculated according to a preset cross-correlation function. The calculation method is as follows:

[0179]

[0180] Where Corr(x(t),y(t+τ)) represents the correlation value between the noise signal x(t) and the vibration signal y(t+τ) at time t; Cov(x(t),y(t+τ)) represents the covariance between x(t) and y(t+τ); δ(x(t)) represents the variance of the noise signal x(t) at time t, and δ(y(t+τ)) represents the variance of the vibration signal y(t+τ) at time t;

[0181] The maximum correlation value is determined as the correlation lag, and the calculation method is as follows:

[0182] τ=argmax(Corr(x(t),y(t+τ)))

[0183] Where τ represents the correlation time delay between the filtered noise signal and the vibration signal.

[0184] In this invention, the process of training the time delay compensation neural network model using correlation time delay includes: taking the existing filtered noise signal and vibration signal as input, estimating the correlation time delay τ between the filtered noise signal and vibration signal as the target output, and then training.

[0185] In this invention, the trained time-delay compensation neural network model is generated based on an LSTM model and a GRU model, such as... Figure 2 As shown.

[0186] In this invention, the method for obtaining a time-delay-compensated vibration signal by performing time-delay compensation using a trained time-delay compensation neural network model based on the estimated time delay between the filtered noise signal and the vibration signal is as follows: the estimated time delay τ(f) between the filtered noise signal and the vibration signal is input into the LSTM model of the trained time-delay compensation neural network model for first time-delay compensation to obtain a first compensated vibration signal; the estimated time delay τ(f) between the filtered noise signal and the vibration signal is input into the GRU model of the trained time-delay compensation neural network model for second time-delay compensation to obtain a second compensated vibration signal; the first compensated vibration signal and the second compensated vibration signal are assigned the same weight value respectively; and the time-delay-compensated vibration signal is obtained based on the first compensated vibration signal, the second compensated vibration signal, the weight value of the first compensated vibration signal, and the weight value of the second compensated vibration signal.

[0187] In a preferred embodiment, a noise suppression signal can be obtained by using a multi-channel feedback system to calculate multi-frequency, multi-channel feedback control signals based on the vibration signal after time delay compensation.

[0188] Furthermore, the dynamic feedback compensation algorithm is based on a multi-channel feedback system, with each channel processing noise in different frequency bands. The method for calculating the noise-suppressed signal using the dynamic feedback compensation algorithm on the time-delay-compensated vibration signal is as follows:

[0189]

[0190] Among them, u k Indicates noise suppression signal; x i (t) represents the vibration signal after time delay compensation from the input of the i-th channel; h i This represents the filter coefficients after the filter settings for the i-th channel.

[0191] Furthermore, this invention dynamically adjusts the feedback gain based on the noise signal frequency and time delay characteristics, optimizing the control parameters of the dynamic feedback compensation algorithm to ensure more stable calculations. The adjustment strategy is as follows:

[0192]

[0193] Where μ(k) represents the feedback gain; α represents the adjustment factor (generally taken between 0 and 1 based on experience).

[0194] Furthermore, the vibration of the noise source can be adjusted in real time through a dynamic feedback loop. Based on the time delay relationship between the vibration signal and the noise source, the noise suppression signal can be adjusted in a timely manner, thereby ensuring the effective suppression of multi-frequency noise.

[0195] After obtaining the noise suppression signal, the driver can be controlled by the noise suppression signal to generate a drive signal. The vibration signal generated by the inertial actuator is driven by the drive signal to actively suppress the noise of the transformer. In a specific implementation, the vibration signal is opposite to the vibration direction of the transformer.

[0196] Specifically, the inertial actuator of the present invention is as follows: Figure 3 As shown, the inertial actuator consists of a mass block, a spring, a frame (shell), and a motor. The mass block is connected to the spring, the motor's mover is connected to the mass block, and the motor's stator is connected to the frame. The mass block, made of metal, is used to generate inertial force. The spring provides elastic support and is used to adjust the system's natural frequency. The motor generates active actuation force to drive the mass block to move, thereby counteracting the vibration of the transformer shell.

[0197] Therefore, when actively suppressing noise in a transformer by driving an inertial actuator (e.g., a piezoelectric actuator or a magnetostrictive actuator) to generate vibration signals based on a drive signal, the motor in the inertial actuator can be driven to generate an actuating force based on the drive signal; then, based on the actuating force, the mass block in the inertial actuator is driven by a spring to generate a vibration signal; finally, the transformer can be actively suppressed based on the vibration signal.

[0198] In some specific embodiments of the present invention, the above-mentioned active noise suppression operation can be implemented in application scenarios such as the active noise suppression device 10. The active noise suppression device 10 can be used for active noise suppression in substations, transformers, etc., and can be, for example, a personal computer (PC), server, or laptop. Figure 4 As shown, the noise active suppression device 10 may include one or more processors 101, memory 102, I / O interface 103, and database 104. Specifically, the processor 101 may be a central processing unit (CPU) or a digital processing unit, etc. The memory 102 may be volatile memory, such as random-access memory (RAM); the memory 102 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or the memory 102 may be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory 102 may be a combination of the above-mentioned memories. The memory 102 may store some program instructions for the noise active suppression operation provided by the present invention. When these program instructions are executed by the processor 101, they can be used to implement the steps of the noise active suppression operation provided by the present invention to effectively suppress noise from substations and transformers. Database 104 can be used to store data such as multi-frequency transformer noise, noise from various noise sources, vibration signals, trained time-delay compensation neural network models, driving signals, vibration signals, frequency domain block minimum mean square FBLMS algorithm, and cross-spectral density values ​​involved in the solution provided by the present invention.

[0199] In this invention, the active noise suppression device 10 can acquire noise acquisition commands through the I / O interface 103. Then, the processor 101 of the active noise suppression device 10 will effectively suppress the noise of the substation and transformer according to the noise active suppression operation program instructions provided by this invention in the memory 102. In addition, the noise of the multi-frequency transformer, the noise of each noise source, vibration signals, the trained time delay compensation neural network model, drive signals, vibration signals, frequency domain block least mean square FBLMS algorithm, and cross-spectral density values ​​can be stored in the database 104.

[0200] The technical solution of the present invention is not limited to Figure 4 The application scenarios shown can also be used in other possible application scenarios.

[0201] In a specific implementation, the hardware system for the active noise suppression operation described above in this invention can be as follows: Figure 5 As shown, the system includes a signal acquisition unit, an adaptive filter, a time delay compensation module, a multi-frequency dynamic feedback controller, multiple drivers, multiple inertial actuators, and a transformer.

[0202] Furthermore, when performing noise suppression based on this active noise suppression operation, firstly, a signal acquisition device can be used to acquire noise from the transformer to obtain a multi-frequency transformer noise signal. Then, an adaptive filter can be used to adaptively filter the multi-frequency transformer noise signal to obtain a filtered multi-frequency transformer noise signal. Next, based on the filtered multi-frequency transformer noise signal, a multi-frequency dynamic feedback controller can be used to estimate the time delay and determine the time delay between the signal source and the vibration signal. Then, based on this time delay, a time delay compensation module can be used to compensate for the time delay to obtain a compensated vibration signal. Then, the compensated vibration signal can be input into the multi-frequency dynamic feedback controller to obtain a multi-dimensional noise suppression signal. Next, based on this multi-dimensional noise suppression, multiple drivers are controlled to generate corresponding drive signals. Then, according to these drive signals, multiple inertial actuators can generate vibration signals to actively suppress noise from the transformer.

[0203] In one possible implementation, a real-time embedded controller (such as a hardware platform based on DSP, FPGA, or ARM architecture) or a programmable logic controller (PLC) system can be used to process noise data, calculate control signals, and drive actuators. These controllers support high-speed data processing and can receive signals from noise and vibration sensors in real time and convert them into control signals.

[0204] In some possible implementations, various aspects of the invention may also be implemented as a program component comprising program code that, when run on a computer device, causes the computer device to perform the steps of the methods described above according to various exemplary embodiments of the present application, such as the computer device performing the operations described above.

[0205] The implementation of all or part of the above steps can be accomplished by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps included in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. Alternatively, if the integrated unit of the present invention is implemented as a software functional module and sold or used as an independent part, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software part. This computer software part is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0206] In summary, the present invention has the following advantages:

[0207] By introducing a well-trained time-delay compensation neural network model and adaptive filtering for time-delay compensation, the present invention accurately estimates and compensates for the time delay between the noise source and the vibration signal. Therefore, compared with the prior art, the present invention not only solves the problems of delay and inaccurate noise suppression caused by time delay, but also significantly improves the system response speed and stability, and optimizes the real-time performance and accuracy of noise suppression.

[0208] Because this invention can suppress noise in different frequency bands in real time, compared with the prior art, it can more accurately adapt to the characteristics of the noise when dealing with multi-frequency noise. It overcomes the shortcomings of the prior art in terms of time delay and noise frequency band processing, and achieves a more efficient and stable noise suppression effect. It has strong real-time optimization capability and low implementation cost, and can significantly improve the overall performance of substation noise control.

[0209] Because this invention can continuously adjust control parameters based on real-time noise signals to ensure the system always maintains optimal operating conditions, it avoids manual intervention. Therefore, compared to existing passive sound insulation methods (such as soundproof enclosures and sound barriers), this invention not only significantly reduces implementation costs by not relying on large physical structures and complex sound insulation projects, but also can be directly applied to existing facilities through the integration of a real-time control system, reducing infrastructure modifications and lowering initial investment and maintenance costs.

[0210] Example 3

[0211] A computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of Embodiment 2.

[0212] It should be understood that any parts not described in detail in this specification belong to the prior art.

[0213] The preferred embodiments of the present invention have been described in detail above; however, the present invention is not limited thereto. Within the scope of the inventive concept, various simple modifications can be made to the technical solutions of the present invention, including combinations of various technical features in any other suitable manner. These simple modifications and combinations should also be considered as the content disclosed in the present invention and are all within the protection scope of the present invention.

Claims

1. An active noise suppression system, characterized in that, include: The time delay estimation module is used to collect noise and vibration signals from noise sources in substations, perform spectrum analysis on the collected noise signals to identify the frequency of the noise sources, filter the noise signals of the corresponding frequency noise sources using a filter with set filtering coefficients to obtain the filtered noise signals, and estimate the time delay between the filtered noise signals and vibration signals using cross-spectral analysis. The noise suppression calculation module is used to perform time delay compensation based on the estimated time delay between the filtered noise signal and the vibration signal, using a trained time delay compensation neural network model to obtain the time delay compensated vibration signal. The dynamic feedback compensation algorithm is then used to calculate the noise suppression signal from the time delay compensated vibration signal.

2. The active noise suppression system according to claim 1, characterized in that, The method for identifying the frequency of the noise source by performing spectral analysis on the collected noise signal is as follows: Where X(f) represents the spectrum of the noise signal; x(n) represents the noise signal corresponding to the nth sampling point in the time domain; f represents the frequency of the noise; N represents the length of the noise signal; and j represents the complex operator.

3. The active noise suppression system according to claim 1, characterized in that, The method for setting the filter coefficients is as follows: The minimum mean square error adaptive filtering method adjusts the filter coefficients step by step according to the error between the input signal and the desired output. Its update formula is as follows: w(k+1)=w(k)+μe(k)x(k) Where w(k) represents the filter coefficient at time k; w(k+1) represents the filter coefficient at time k+1; μ represents the filter step size parameter; e(k) represents the error signal between the input signal and the desired output at time k; and x(k) represents the input signal at time k.

4. The active noise suppression system according to claim 1 or 3, characterized in that, The method for estimating the time delay between the filtered noise signal and the vibration signal using cross-spectral analysis is as follows: The cross-spectral density value between the noise signal and the vibration signal of the noise source is calculated using the following method: Among them, S xy (f) represents the cross-spectral density values ​​of x(n) and y(n); y * (n) represents the conjugate complex number of y(n); y(n) represents the vibration signal corresponding to the nth sampling point in the time domain; Based on the cross-spectral density values ​​between the noise signal and the vibration signal from the noise source, the time delay τ(f) between the filtered noise signal and the vibration signal is estimated using the following method: Among them, ∠S xy (f) represents the cross-spectral density value S xy (f) corresponds to the phase value.

5. The active noise suppression system according to claim 1, characterized in that, The correlation time delay between the filtered noise signal and the vibration signal is estimated by using correlation analysis. The correlation time delay is then used to train the time delay compensation neural network model, resulting in the trained time delay compensation neural network model. The method for estimating the correlation time delay between the filtered noise signal and the vibration signal using correlation analysis is as follows: The correlation value between the noise signal and the vibration signal of the noise source is calculated using the following method: Where Corr(x(t),y(t+τ)) represents the correlation value between the noise signal x(t) and the vibration signal y(t+τ) at time t; Cov(x(t),y(t+τ)) represents the covariance between x(t) and y(t+τ); δ(x(t)) represents the variance of the noise signal x(t) at time t, and δ(y(t+τ)) represents the variance of the vibration signal y(t+τ) at time t; The maximum correlation value is determined as the correlation lag, and the calculation method is as follows: τ=argmax(Corr(x(t),y(t+τ))) Where τ represents the correlation time delay between the filtered noise signal and the vibration signal.

6. The active noise suppression system according to claim 1 or 5, characterized in that, The trained time delay compensation neural network model is generated based on the interaction of the LSTM model and the GRU model. Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using a trained time delay compensation neural network model to obtain the time delay-compensated vibration signal. The method is as follows: the estimated time delay between the filtered noise signal and the vibration signal is input into the LSTM model of the trained time delay compensation neural network model for first time delay compensation to obtain a first compensated vibration signal; the estimated time delay between the filtered noise signal and the vibration signal is input into the GRU model of the trained time delay compensation neural network model for second time delay compensation to obtain a second compensated vibration signal; weight values ​​are assigned to the first compensated vibration signal and the second compensated vibration signal respectively; and the time delay-compensated vibration signal is obtained based on the first compensated vibration signal, the second compensated vibration signal, the weight value of the first compensated vibration signal, and the weight value of the second compensated vibration signal.

7. The active noise suppression system according to claim 1, characterized in that, The dynamic feedback compensation algorithm is based on a multi-channel feedback system, with each channel processing noise in different frequency bands; The method for calculating the noise-suppressed signal using a dynamic feedback compensation algorithm on the vibration signal after time-delay compensation is as follows: Among them, u k Indicates noise suppression signal; x i (t) represents the vibration signal after time delay compensation from the input of the i-th channel; h i This represents the filter coefficients after the filter settings for the i-th channel; Simultaneously, based on the noise signal frequency and time delay characteristics, the feedback gain is dynamically adjusted to optimize the control parameters of the dynamic feedback compensation algorithm, as follows: Where μ(k) represents the feedback gain and α represents the adjustment factor.

8. A method for active noise suppression, characterized in that, include: Noise and vibration signals from noise sources in a substation are collected. The collected noise signals are subjected to spectrum analysis to identify the frequency of the noise source. The noise signals of the corresponding frequency noise sources are filtered using a filter with set filtering coefficients to obtain the filtered noise signals. Cross-spectrum analysis is used to estimate the time delay between the filtered noise signals and the vibration signals. Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using the trained time delay compensation neural network model to obtain the time delay compensated vibration signal. The noise suppression signal is then calculated using a dynamic feedback compensation algorithm.

9. The active noise suppression method according to claim 8, characterized in that, The method for identifying the frequency of the noise source by performing spectral analysis on the collected noise signal is as follows: Where X(f) represents the spectrum of the noise signal; x(n) represents the noise signal corresponding to the nth sampling point in the time domain; f represents the frequency of the noise; N represents the length of the noise signal; and j represents the complex operator.

10. The active noise suppression method according to claim 8, characterized in that, The method for setting the filter coefficients is as follows: The minimum mean square error adaptive filtering method adjusts the filter coefficients step by step according to the error between the input signal and the desired output. Its update formula is as follows: w(k+1)=w(k)+μe(k)x(k) Where w(k) represents the filter coefficient at time k; w(k+1) represents the filter coefficient at time k+1; μ represents the filter step size parameter; e(k) represents the error signal between the input signal and the desired output at time k; and x(k) represents the input signal at time k.

11. The active noise suppression method according to claim 8 or 10, characterized in that, The method for estimating the time delay between the filtered noise signal and the vibration signal using cross-spectral analysis is as follows: The cross-spectral density value between the noise signal and the vibration signal of the noise source is calculated using the following method: Among them, S xy (f) represents the cross-spectral density values ​​of x(n) and y(n); y * (n) represents the conjugate complex number of y(n); y(n) represents the vibration signal corresponding to the nth sampling point in the time domain; Based on the cross-spectral density values ​​between the noise signal and the vibration signal from the noise source, the time delay τ(f) between the filtered noise signal and the vibration signal is estimated using the following method: Among them, ∠S xy (f) represents the cross-spectral density value S xy (f) corresponds to the phase value.

12. The active noise suppression method according to claim 8, characterized in that, The correlation time delay between the filtered noise signal and the vibration signal is estimated by using correlation analysis. The correlation time delay is then used to train the time delay compensation neural network model, resulting in the trained time delay compensation neural network model. The method for estimating the correlation time delay between the filtered noise signal and the vibration signal using correlation analysis is as follows: The correlation value between the noise signal and the vibration signal of the noise source is calculated using the following method: Where Corr(x(t),y(t+τ)) represents the correlation value between the noise signal x(t) and the vibration signal y(t+τ) at time t; Cov(x(t),y(t+τ)) represents the covariance between x(t) and y(t+τ); δ(x(t)) represents the variance of the noise signal x(t) at time t, and δ(y(t+τ)) represents the variance of the vibration signal y(t+τ) at time t; The maximum correlation value is determined as the correlation lag, and the calculation method is as follows: τ=argmax(Corr(x(t),y(t+τ))) Where τ represents the correlation time delay between the filtered noise signal and the vibration signal.

13. The active noise suppression method according to claim 8 or 12, characterized in that, The trained time delay compensation neural network model is generated based on the interaction of the LSTM model and the GRU model. Based on the estimated time delay between the filtered noise signal and the vibration signal, the time delay compensation is performed using a trained time delay compensation neural network model to obtain the time delay-compensated vibration signal. The method is as follows: the estimated time delay between the filtered noise signal and the vibration signal is input into the LSTM model of the trained time delay compensation neural network model for first time delay compensation to obtain a first compensated vibration signal; the estimated time delay between the filtered noise signal and the vibration signal is input into the GRU model of the trained time delay compensation neural network model for second time delay compensation to obtain a second compensated vibration signal; weight values ​​are assigned to the first compensated vibration signal and the second compensated vibration signal respectively; and the time delay-compensated vibration signal is obtained based on the first compensated vibration signal, the second compensated vibration signal, the weight value of the first compensated vibration signal, and the weight value of the second compensated vibration signal.

14. The active noise suppression method according to claim 8, characterized in that, The dynamic feedback compensation algorithm is based on a multi-channel feedback system, with each channel processing noise in different frequency bands; The method for calculating the noise-suppressed signal using a dynamic feedback compensation algorithm on the vibration signal after time-delay compensation is as follows: Among them, u k Indicates noise suppression signal; x i (t) represents the vibration signal after time delay compensation from the input of the i-th channel; h i This represents the filter coefficients after the filter settings for the i-th channel; Simultaneously, based on the noise signal frequency and time delay characteristics, the feedback gain is dynamically adjusted to optimize the control parameters of the dynamic feedback compensation algorithm, as follows: Where μ(k) represents the feedback gain and α represents the adjustment factor.

15. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 8-14.

Images