A transformer sound signal processing method and system based on an improved ICA algorithm

Abstract

The application discloses a transformer sound signal processing method and system based on an improved ICA algorithm, and the method comprises the following steps: inputting a noisy transformer sound signal to obtain a mixed to-be-processed signal matrix; performing decomposition on the underdetermined mixed to-be-processed audio signal matrix based on an EEMD algorithm improved based on a minimum lower limit frequency to obtain IMF components and perform preliminary denoising; estimating the dimension of a source signal based on a sound print dimension recognition model of a GRU network, upgrading the dimension of the sound signal matrix, and judging whether the upgraded signal matrix is underdetermined; if underdetermined, extracting a residual error; if not underdetermined, screening an upgraded signal group based on a maximum correlation principle, and performing blind source signal decomposition based on an N-MNE improved ICA algorithm. The method can effectively improve the processing effect of the transformer sound signal.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power systems, and particularly relates to a transformer sound signal processing method and system based on an improved ICA algorithm. BACKGROUND

[0002] The transformer is a key device in the high-voltage and long-distance power transmission system in the transformer substation. At the same time, the transformer is also one of the main noise sources in the transformer substation. In the operation process, the sound signals generated by the mechanical vibration of the transformer winding, the iron core, the oil tank and other structures contain rich state characteristics, the fault of the transformer is closely related to the voiceprint, and the abnormality of the transformer body sound signal is highly related to the fault of the transformer itself. Moreover, compared with the power-off detection such as the transition resistance method, the voiceprint recognition has become a research hotspot of domestic and foreign scholars in recent years due to its real-time performance, convenience and economy. It is of great significance to develop the transformer voiceprint detection and fault diagnosis to ensure the safe operation of the transformer. The microphone array is designed to collect the sound signals of the transformer, and the working condition detection and diagnosis of the transformer can be realized through the analysis of the sound signals. However, there are a large number of interference noise signals such as ambient noise, corona discharge, bird chirping sound and the like in the process of collecting the sound signals of the power transformer, which leads to the fact that the pure sound signals of the working state of the transformer cannot be obtained efficiently and quickly, and the fault state monitoring system often makes a wrong judgment.

[0003] Blind source separation (BSS) is a technology for recovering source signals only by observing signals without prior knowledge of source signals and mixing coefficients. Due to its advancement and superiority, it has become a research hotspot of domestic and foreign scholars in recent years. When the number of microphone array elements is less than the number of noise source signals, the BSS problem is called underdetermined blind source separation (UBSS), and due to its more extensive application occasions, UBSS has become a research hotspot in the field of signal processing internationally. The FastICA (Fast Independent Component Analysis) algorithm is a common algorithm for solving the problem of blind source separation, and has good separation precision under non-underdetermined conditions. However, in actual working conditions, the types of noise sources are generally random and complex (up to dozens of types), and the number of noise sensors is underdetermined. How to meet the reversibility constraint of the mixing process and the uniqueness constraint of the separated signals of the FastICA algorithm in the blind source separation of the transformer voiceprint becomes a key problem in solving the underdetermined blind source separation of the transformer voiceprint. SUMMARY

[0004] The application provides a transformer sound signal processing method and system based on an improved ICA algorithm, which is used to solve the technical problem of underdetermined blind source separation of the transformer voiceprint.

[0005] In a first aspect, the application provides

[0006] In a second aspect, the application provides

[0007] In a third aspect, an electronic device is provided, comprising at least one processor, and a memory connected with the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the transformer acoustic signal processing method based on the improved ICA algorithm of any of the embodiments of the present application.

[0008] In a fourth aspect, the present application further provides a computer readable storage medium having stored thereon a computer program, and the program instructions are executed by a processor to enable the processor to perform the steps of the transformer acoustic signal processing method based on the improved ICA algorithm of any of the embodiments of the present application.

[0009] The transformer acoustic signal processing method and system based on the improved ICA algorithm of the present application, input the noisy transformer acoustic signal to obtain a mixed to-be-processed signal matrix; the underdetermined mixed to-be-processed audio signal matrix is decomposed to obtain IMF components and preliminarily denoised based on the EEMD algorithm improved based on the minimum lower limit frequency, the source signal dimension is estimated based on the voiceprint dimension recognition model of the GRU network, the dimension of the upgraded acoustic signal matrix is upgraded, and it is judged whether the upgraded signal matrix is underdetermined, if underdetermined, the residual error is extracted, if not underdetermined, the upgraded signal group is screened based on the maximum correlation principle, and the component blind source signal is decomposed based on the N-MNE improved ICA algorithm, which can effectively improve the processing effect of the transformer acoustic signal. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0011] Figure 1 A flow chart of a transformer acoustic signal processing method based on an improved ICA algorithm provided by an embodiment of the present application is provided.

[0012] Figure 2 A structural block diagram of a transformer acoustic signal processing system based on an improved ICA algorithm provided by an embodiment of the present application is provided.

[0013] Figure 3 A structural schematic diagram of an electronic device provided by an embodiment of the present application is provided. DETAILED DESCRIPTION

[0014] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0015] Referring to Figure 1 , a flowchart of a transformer acoustic signal processing method based on an improved ICA algorithm is shown.

[0016] As Figure 1 shown, the transformer acoustic signal processing method based on the improved ICA algorithm specifically comprises the following steps:

[0017] Step S101, constructing a mixed to-be-processed signal matrix according to the obtained noise-containing transformer signal.

[0018] Step S102, decomposing the mixed to-be-processed signal matrix by using an improved EEMD algorithm based on a minimum lower limit frequency to obtain at least one IMF component, and preliminarily denoising the at least one IMF component to obtain at least one target IMF component.

[0019] In this step, the noise-containing transformer acoustic signal collected by the i th microphone is subjected to short-time Fourier transform power spectrum analysis, and the minimum frequency in the noise-containing transformer acoustic signal is found out as the cutoff frequency f L of signal decomposition corresponding to the noise-containing transformer acoustic signal, and is taken as the lower limit of signal decomposition.

[0020] In the noise-containing transformer signal, a standard Gaussian white noise is added to obtain a mixed noise signal, and the expression is:

[0021] M i (t)=m i (t)+n w (t),

[0022] In the formula, m i (t) is the i th noise-containing transformer acoustic signal, n w (t) is a standard Gaussian white noise, and M i (t) is the i th mixed noise signal.

[0023] The mixed noise signal is subjected to EEMD decomposition to obtain an IMF component and a residual error, and the expression is:

[0024]

[0025] In the formula, I i，j(t) is the jth order IMF component of the ith noisy transformer sound signal, r i (t) is the residual of the EEMD decomposition of the ith noisy transformer sound signal, J is the order number of the IMF component group;

[0026] The I i，j (t) is decomposed into short-time Fourier transform power spectrum analysis, and the target frequency f i，j (t) is obtained. max The target frequency f max is compared with the cutoff frequency f L , if the target frequency f max is less than the cutoff frequency f L , the decomposition is ended, otherwise, the decomposition is continued, and finally the N order IMF component of the EEMD decomposed ith noisy transformer sound signal is obtained.

[0027] The average of each IMF component is taken, and after the influence of Gaussian white noise is offset, the overall average of each order IMF component after the decomposition of the noisy transformer sound signal is obtained, and the expression is:

[0028]

[0029] In the formula, C k，j is the overall average of the jth order IMF component of K times of decomposition, C i，j is the jth order IMF component of the ith noisy transformer sound signal, and M is the number of noisy transformer sound signals.

[0030] The residual r(t) of the noisy transformer sound signal after EEMD decomposition is calculated, and the expression is:

[0031]

[0032] The correlation coefficient of each order IMF component and the corresponding noisy transformer sound signal is calculated, and the expression is:

[0033]

[0034] In the formula, ρ k is the correlation coefficient value of the kth order IMF component and the corresponding noisy transformer sound signal, C k (t) is the kth order IMF component, C abs is the average value of the kth order IMF component, x k (t) is the transformer sound signal, and x abs is the average value of the transformer sound signal, and K' is the order number of the IMF component.

[0035] A threshold μ is set, and the IMF components with correlation coefficient values greater than the threshold μ are reserved to obtain at least one target IMF component, and the expression of the threshold μ is:

[0036]

[0037] In the formula, is the average value of p k .

[0038] It should be noted that after the mixed to-be-processed signal matrix is decomposed by using the improved EEMD algorithm based on the minimum lower limit frequency to obtain at least one IMF component, a voiceprint degree recognition model based on a GRU network is constructed, the dimension of the noise-containing transformer signal is recognized according to the voiceprint degree recognition model, and it is judged whether the dimension of the noise-containing transformer signal is greater than the dimension of a certain IMF component, wherein the expression for calculating the dimension of the noise-containing transformer signal is:

[0039] S e = GRU(m i (t)),

[0040] In the formula, m i (t) is the i th noise-containing transformer acoustic signal, GRU is a gated recurrent unit, and S e is the dimension of the noise-containing transformer signal.

[0041] If it is greater than the dimension of a certain IMF component, the residual corresponding to the certain IMF component is decomposed again based on the improved EEMD algorithm.

[0042] If it is not greater than the dimension of a certain IMF component, the certain IMF component is subjected to preliminary denoising.

[0043] Specifically, a transformer voiceprint simulation laboratory is used to simulate on-site transformer noise-free fault voiceprints, single-interference-source fault voiceprints, two-interference-source fault voiceprints, and more than two noise interference source fault voiceprints, and a voiceprint dimension recognition model database is constructed based on the simulated voiceprints.

[0044] Since the transformer acoustic signal has strong time sequence characteristics, a recurrent neural network that is good at analyzing the time sequence relationship between discrete data sequences is considered to be used. The GRU network is a simplified long short-term memory (LSTM) network that combines the input gate and the forgetting gate in the LSTM network into an update gate to reduce the number of parameters and alleviate overfitting, while improving the calculation efficiency and maintaining a prediction effect close to that of the LSTM. The process of constructing the voiceprint dimension recognition model includes:

[0045] The voiceprint dimension recognition model based on the GRU network is trained using a voiceprint dimension recognition model database, in the model training stage, a voiceprint dimension recognition model and a voiceprint dimension prediction model are respectively built, the hyperparameters of training are set, including batch size, learning rate and iteration number, etc., the recognition accuracy of the model and the prediction error of the prediction model are verified.

[0046] The reset gate control signal r of the GRU network t The expression of r' is:

[0047] r' t = Sig(W r X t + U r h t-1 ),

[0048] In the formula, W r is a reset gate connection weight matrix, X t is the current input, h t-1 is the hidden gate output at t-1 time, U r is a reset gate and previous time step hidden state connection weight matrix, which extracts features related to the reset gate from the previous time step hidden state, and Sig is a Sigmoid function.

[0049] The update gate control signal Z of the GRU network t The expression of Z is:

[0050] Z t = Sig{W Z [h t-1 , x t ]},

[0051] In the formula, W Z is an update gate connection weight matrix.

[0052] c t represents the memory information at t time, and the expression is:

[0053] c t = tanh(W c X t + U c (r t ' o h t-1 )),

[0054] In the formula, W c and U c both represent learnable weight parameters, and o represents dot product.

[0055] h t is the hidden gate output at t time, and the expression is:

[0056] ht = z t h t-1 + (1 - z t ) c t .

[0057] Step S103, based on the maximum correlation principle, screening the first S e order target IMF components with the maximum correlation with the noisy transformer signal to form an upgraded signal group D u , wherein S e is the dimension of the noisy transformer signal.

[0058] In this step, the expression for calculating the correlation is:

[0059]

[0060] In the formula, cov(·) is the covariance function, σ(·) is the standard deviation function, is the correlation coefficient, m(t) is the noisy transformer signal, C i，j (t) is the jth IMF component of the ith noisy transformer signal;

[0061] The upgraded signal group is equivalent to the original underdetermined voiceprint signal group after linear transformation. Since there is no nonlinear process, using the IMF signal group as input for blind source separation will not destroy the linear requirement of the FastICA algorithm, and the upgraded signal group solves the key problems in the underdetermined blind source separation of the transformer voiceprint signal, such as the reversibility constraint of the mixed signal and the uniqueness constraint of the separated signal.

[0062] Further, the expression of the upgraded signal group D u is:

[0063]

[0064] In the formula, C1(t) is the first IMF component with the maximum correlation with the noisy transformer signal, C2(t) is the second IMF component with the maximum correlation with the noisy transformer signal, is the S e th IMF component with the maximum correlation with the noisy transformer signal.

[0065] Step S104, based on the preset N-MNE-ICA model, performing blind source separation on the upgraded signal group D u to obtain a blind source signal.

[0066] In this step, due to the difficulty of traditional negative entropy calculation method, the accuracy of empirical formula is not high enough, so it is necessary to improve it, that is, the independent component analysis algorithm (N-MNE-ICA) based on Newton limit-maximum information negative entropy (N-MNE) improvement. The signal group after dimensionality increase is input into the ICA algorithm module based on N-MNE improvement, so as to reduce the calculation complexity of negative entropy, improve the convergence speed and increase the separation accuracy.

[0067] The maximum entropy principle assumes that the information is not made without any unknown assumption, and the unknown event is regarded as an equal probability event. By substituting the maximum entropy principle into the negative entropy formula, the maximum approximate negative entropy can be obtained, and the expression of the maximum approximate negative entropy is:

[0068]

[0069] In the formula, J max (y) is the maximum approximate negative entropy, N and E are the total number of samples and mathematical expectation respectively, k i is a weight factor, which is a constant, G i is a non-quadratic function, y and v are variables with mean 0 and variance 1;

[0070] In order to obtain the maximum approximate negative entropy value of y=B T x, the non-Gaussianity of y is maximized by E{G i (y)}, and the expression of the constraint condition is:

[0071]

[0072] In the formula, B is the convergence threshold function, B T is the transpose of the convergence threshold function B, χ is the input signal, J G is the input signal, G′ i is the derivative of the non-quadratic function, J G is a non-Gaussianity measure function based on the input signal, and β′ is a Lagrange multiplier operator used to constrain the optimization problem.

[0073] The maximum entropy theory and the negative entropy formula have proved that there is an optimal solution, which can make the constraint equation equal. Assuming that the optimal solution is B0, the constraint condition is solved to obtain:

[0074]

[0075] In the formula, is the transpose of the optimal solution;

[0076] According to Newton's limit theorem, the tangent line to a curve can be used to approximate the curve, and therefore is also approximately the root of the tangent line. If the required accuracy is not met, Newton's method continues until convergence. The general formula for Newton's limit method is:

[0077]

[0078] In the formula, x n Let x be the estimated value for the nth iteration. n+1 f(x) is the estimated value for the (n+1)th iteration. n f'(x) is the Newton limit function. n ) is the derivative of the Newton limit function.

[0079] The iterative formula for Newton's limit method is:

[0080]

[0081] In the formula, B * G" is an iterative function i Let x be the second derivative of a non-quadratic function, and x be the input signal.

[0082] The application of Newton's limit method not only simplifies complex calculations such as solving probability density functions, but also reduces the accumulation of errors that may be caused by high-order calculations in empirical formulas. This method transforms the originally complex statistical mathematical structure into a relatively simple iterative root-finding structure, thereby improving the accuracy and efficiency of the calculation.

[0083] To facilitate subsequent blind source separation calculations, a de-averaging process should be performed before the calculation. After de-averaging, the data is distributed around the coordinate axes, and the function obtained through random initialization can more quickly approximate the target function. The de-averaging expression is:

[0084]

[0085] In the formula, Z is the transformer acoustic signature sample after de-averaging, α is the feature vector of the transformer acoustic signature sample X, and λ is the feature value of X.

[0086] Set the iteration function B * The expression is:

[0087] B * =E{ZG′ i (B T Z)}-E{G″ i (B T Z)}B,

[0088] Define the convergence threshold function B, with the expression:

[0089]

[0090] It should be noted that in the process of blind source separation of the ascending order signal group D u based on the preset N-MNE-ICA model, the signal similarity coefficient is used as an evaluation index to evaluate the separation result, and the expression for calculating the signal similarity coefficient is:

[0091]

[0092] In the formula, ξ ij is the signal similarity coefficient of the i-th signal in the alpha group and the j-th signal in the beta group, ξ ij = [0, 1], when ξ ij = 1, it is considered that the i-th signal is completely consistent with the j-th signal, and ξ ij = 0 is the opposite, alpha i (t) is the i-th signal in the alpha group, beta j (t) is the j-th signal in the beta group, and ξ(alpha i , beta j ) is the signal similarity coefficient of alpha i (t) and beta j (t), and N' is the upper limit of the sound signal time t.

[0093] In summary, the method of the present application inputs the noise-containing transformer sound signal, obtains a mixed to-be-processed signal matrix, decomposes the underdetermined mixed to-be-processed audio signal matrix to obtain IMF components and performs preliminary denoising based on the EEMD algorithm improved based on the minimum lower limit frequency, estimates the source signal dimension based on the voiceprint dimension recognition model of the GRU network, upgrades the dimension of the sound signal matrix, and judges whether the upgraded signal matrix is underdetermined. If it is underdetermined, the residual error is extracted, and if it is not underdetermined, the upgraded signal group is screened based on the maximum correlation principle, and the component blind source signal is separated based on the N-MNE improved ICA algorithm, which can effectively improve the processing effect of the transformer sound signal.

[0094] Please refer to Figure 2 , which shows the structure block diagram of a transformer sound signal processing system based on an improved ICA algorithm.

[0095] As Figure 2 shown, the transformer sound signal processing system 200 includes a construction module 210, a decomposition module 220, a screening module 230, and a separation module 240.

[0096] The constructing module 210 is configured to construct a mixed to-be-processed signal matrix according to the obtained noise-containing transformer signal; the decomposing module 220 is configured to decompose the mixed to-be-processed signal matrix by using an improved EEMD algorithm based on a minimum lower limit frequency to obtain at least one IMF component, and to preliminarily denoise the at least one IMF component to obtain at least one target IMF component; the screening module 230 is configured to screen, based on a maximum correlation principle, the first S e order target IMF components to form an up-sampled signal group D u , wherein S e is a dimension of the noise-containing transformer signal; the separating module 240 is configured to perform blind source separation on the up-sampled signal group D u based on a preset N-MNE-ICA model to obtain a blind source signal.

[0097] It should be understood that Figure 2 the modules described in the above Figure 1 correspond to the respective steps in the methods described with reference to Figure 2 . Thus, the operations and features described above for the methods and the corresponding technical effects apply equally to the modules in , and will not be described again here.

[0098] In some other embodiments, the present application also provides a computer readable storage medium having stored thereon a computer program, the program instructions being executed by a processor to cause the processor to perform the transformer acoustic signal processing method based on the improved ICA algorithm in any of the method embodiments described above.

[0099] As an implementation form, the computer readable storage medium of the present application stores computer executable instructions, which are configured to:

[0100] construct a mixed to-be-processed signal matrix according to the obtained noise-containing transformer signal;

[0101] decompose the mixed to-be-processed signal matrix by using an improved EEMD algorithm based on a minimum lower limit frequency to obtain at least one IMF component, and preliminarily denoise the at least one IMF component to obtain at least one target IMF component;

[0102] screen, based on a maximum correlation principle, the first S e order target IMF components to form an up-sampled signal group D u , wherein S e is a dimension of the noise-containing transformer signal;

[0103] perform blind source separation on the up-sampled signal group D uBlind source separation is performed to obtain blind source signals.

[0104] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the transformer acoustic signal processing system based on the improved ICA algorithm, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely configured relative to a processor, which can be connected to the transformer acoustic signal processing system based on the improved ICA algorithm via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0105] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the transformer acoustic signal processing method based on the improved ICA algorithm described in the above method embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the transformer acoustic signal processing system based on the improved ICA algorithm. The output device 340 may include a display screen or other display device.

[0106] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0107] In one implementation, the above-described electronic device is applied to a transformer acoustic signal processing system based on an improved ICA algorithm, for a client, and includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0108] According to the obtained noise-containing transformer signal, a mixed to-be-processed signal matrix is constructed;

[0109] An improved EEMD algorithm based on a minimum lower limit frequency is adopted to decompose the mixed to-be-processed signal matrix, at least one IMF component is obtained, and the at least one IMF component is preliminarily denoised to obtain at least one target IMF component;

[0110] Based on the maximum correlation principle, the first S e order target IMF components with the greatest correlation to the noise-containing transformer signal are screened out to form an up-sampling signal group D u , wherein S e is the dimension of the noise-containing transformer signal.

[0111] Based on a preset N-MNE-ICA model, blind source separation is performed on the up-sampling signal group D u to obtain a blind source signal.

[0112] Through the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions or the essential part of the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the method of each embodiment or some part of the embodiment.

[0113] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A transformer acoustic signal processing method based on an improved ICA algorithm, characterized in that, include: A hybrid signal matrix to be processed is constructed based on the acquired noisy transformer signal; The hybrid signal matrix to be processed is decomposed using an improved EEMD algorithm based on the minimum lower limit frequency to obtain at least one IMF component, and the at least one IMF component is initially denoised to obtain at least one target IMF component. Based on the principle of maximum correlation, the top results with the highest correlation to the noisy transformer signal were selected. The first-order target IMF components constitute the upgraded signal group ,in, For noisy transformer signals; The raised-order signal group is based on the preset N-MNE-ICA model. Blind source separation is performed to obtain blind source signals. After decomposing the mixed signal matrix to be processed using an improved EEMD algorithm based on the minimum lower limit frequency to obtain at least one IMF component, the method further includes: A voiceprint hierarchy identification model based on a GRU network is constructed. The model is used to identify the dimension of the noisy transformer signal and determine whether the dimension of the noisy transformer signal is greater than the dimension of a certain IMF component. The expression for calculating the dimension of the noisy transformer signal is as follows: ， In the formula, For the i-th noisy transformer acoustic signal, For gated loop unit, For noisy transformer signals; If the dimension is greater than that of a certain IMF component, the residual corresponding to that IMF component will be decomposed again based on the improved EEMD algorithm. If the dimension is not greater than that of a certain IMF component, then the certain IMF component will undergo preliminary denoising processing.

2. The transformer acoustic signal processing method based on the improved ICA algorithm according to claim 1, characterized in that, The decomposition of the mixed signal matrix to be processed using the improved EEMD algorithm based on the minimum lower limit frequency to obtain at least one IMF component includes: Standard Gaussian white noise is added to the noisy transformer signal to obtain a mixed noise signal, expressed as follows: ， In the formula, For the i-th noisy transformer acoustic signal, Standard Gaussian white noise, This is the i-th mixed noise signal; EEMD decomposition is performed on the mixed noise signal to obtain the IMF component and residual, expressed as follows: ， In the formula, The j-th order IMF component is obtained by decomposing the i-th noisy transformer acoustic signal. The residual obtained from the EEMD decomposition of the i-th noisy transformer acoustic signal is... The order of the IMF component group; For the decomposed Short-time Fourier transform power spectrum analysis was performed to obtain Target frequency , target frequency With cutoff frequency In comparison, if the target frequency Less than the cutoff frequency If the result is positive, the decomposition ends; otherwise, the decomposition continues, ultimately yielding the Nth order IMF component of the i-th noisy transformer acoustic signal after EEMD decomposition. After averaging each IMF component to cancel out the Gaussian white noise, the overall average of each order of IMF components after decomposing the noisy transformer acoustic signal is obtained, expressed as: ， In the formula, To calculate the global average of the j-th order IMF component of the K-th decomposition, Let j be the j-th order IMF component of the i-th noisy transformer acoustic signal. This represents the quantity of acoustic signals from a noisy transformer. Calculate the residual obtained after EEMD decomposition of the acoustic signal of a noisy transformer. The expression is: 。 3. The transformer acoustic signal processing method based on the improved ICA algorithm according to claim 2, characterized in that, The preliminary denoising of the at least one IMF component to obtain at least one target IMF component includes: Calculate the correlation coefficients between each order IMF component and the corresponding noisy transformer acoustic signal, as shown in the following expression: ， In the formula, Let be the correlation coefficient value between the k-th order IMF component and the corresponding noisy transformer acoustic signal. For the k-th order IMF component, The average value of the k-th order IMF component. This is a transformer acoustic signal. This represents the average value of the transformer's acoustic signal. The order of the IMF components; Set threshold Retain correlation coefficient values ​​greater than the threshold. IMF components, obtain at least one target IMF component, and calculate the threshold. The expression is: ， In the formula, for The average value.

4. The transformer acoustic signal processing method based on the improved ICA algorithm according to claim 1, characterized in that, The up-order signal group The expression is: ， In the formula, The first IMF component with the highest correlation to the noisy transformer signal. The second-highest correlated IMF component of the noisy transformer signal. The top 100 signals with the highest correlation to noisy transformer signals One IMF component.

5. The transformer acoustic signal processing method based on the improved ICA algorithm according to claim 1, characterized in that, in, Based on the preset N-MNE-ICA model, the upgraded signal group is analyzed. In the process of blind source separation, the signal similarity coefficient is used as an evaluation index to evaluate the separation results. The expression for calculating the signal similarity coefficient is as follows: ， In the formula, for The i-th signal in the group The signal similarity coefficient of the j-th signal in the group. ,when At that time, it is assumed that the i-th signal is completely identical to the j-th signal. Conversely, for The i-th signal in the group, for The j-th signal in the group, for and The signal similarity coefficient, This represents the upper limit of the acoustic signal time t.

6. A transformer acoustic signal processing system based on an improved ICA algorithm, used to execute the method according to any one of claims 1-5, characterized in that, The transformer acoustic signal processing system includes: The module is configured to construct a mixed signal matrix to be processed based on the acquired noisy transformer signal; The decomposition module is configured to decompose the hybrid signal matrix to be processed using an improved EEMD algorithm based on the minimum lower limit frequency to obtain at least one IMF component, and to perform preliminary denoising on the at least one IMF component to obtain at least one target IMF component. The filtering module is configured to filter the signals most correlated with the noisy transformer signal based on the principle of maximum correlation. The first-order target IMF components constitute the upgraded signal group ,in, For noisy transformer signals; The separation module is configured to process the raised-order signal group based on a preset N-MNE-ICA model. Blind source separation is performed to obtain blind source signals.

7. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method described in any one of claims 1 to 5.

Images